poly_int: vector_alignment_reachable_p
[official-gcc.git] / gcc / ChangeLog
blobe42270c40195b3f149472774c5dcaa6b83dcfa8b
1 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
2             Alan Hayward  <alan.hayward@arm.com>
3             David Sherwood  <david.sherwood@arm.com>
5         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
6         number of units as polynomial.
8 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9             Alan Hayward  <alan.hayward@arm.com>
10             David Sherwood  <david.sherwood@arm.com>
12         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13         * target.def (autovectorize_vector_sizes): Return the vector sizes
14         by pointer, using vector_sizes rather than a bitmask.
15         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18         Likewise.
19         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
20         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
21         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
22         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
23         * omp-general.c (omp_max_vf): Likewise.
24         * omp-low.c (omp_clause_aligned_alignment): Likewise.
25         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
26         * tree-vect-loop.c (vect_analyze_loop): Likewise.
27         * tree-vect-slp.c (vect_slp_bb): Likewise.
28         * doc/tm.texi: Regenerate.
29         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
30         to a poly_uint64.
31         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
32         the vector size as a poly_uint64 rather than an unsigned int.
33         (current_vector_size): Change from an unsigned int to a poly_uint64.
34         (get_vectype_for_scalar_type): Update accordingly.
35         * tree.h (build_truth_vector_type): Take the size and number of
36         units as a poly_uint64 rather than an unsigned int.
37         (build_vector_type): Add a temporary overload that takes
38         the number of units as a poly_uint64 rather than an unsigned int.
39         * tree.c (make_vector_type): Likewise.
40         (build_truth_vector_type): Take the number of units as a poly_uint64
41         rather than an unsigned int.
43 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
44             Alan Hayward  <alan.hayward@arm.com>
45             David Sherwood  <david.sherwood@arm.com>
47         * target.def (get_mask_mode): Take the number of units and length
48         as poly_uint64s rather than unsigned ints.
49         * targhooks.h (default_get_mask_mode): Update accordingly.
50         * targhooks.c (default_get_mask_mode): Likewise.
51         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
52         * doc/tm.texi: Regenerate.
54 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
55             Alan Hayward  <alan.hayward@arm.com>
56             David Sherwood  <david.sherwood@arm.com>
58         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
59         * omp-general.c (omp_max_vf): Likewise.
60         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
61         (expand_omp_simd): Handle polynomial safelen.
62         * omp-low.c (omplow_simd_context): Add a default constructor.
63         (omplow_simd_context::max_vf): Change from int to poly_uint64.
64         (lower_rec_simd_input_clauses): Update accordingly.
65         (lower_rec_input_clauses): Likewise.
67 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
68             Alan Hayward  <alan.hayward@arm.com>
69             David Sherwood  <david.sherwood@arm.com>
71         * tree-vectorizer.h (vect_nunits_for_cost): New function.
72         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
73         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
74         (vect_analyze_slp_cost): Likewise.
75         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
76         (vect_model_load_cost): Likewise.
78 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
79             Alan Hayward  <alan.hayward@arm.com>
80             David Sherwood  <david.sherwood@arm.com>
82         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
83         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
84         from an unsigned int * to a poly_uint64_pod *.
85         (calculate_unrolling_factor): New function.
86         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
88 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
89             Alan Hayward  <alan.hayward@arm.com>
90             David Sherwood  <david.sherwood@arm.com>
92         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
93         from an unsigned int to a poly_uint64.
94         (_loop_vec_info::slp_unrolling_factor): Likewise.
95         (_loop_vec_info::vectorization_factor): Change from an int
96         to a poly_uint64.
97         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
98         (vect_get_num_vectors): New function.
99         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
100         (vect_get_num_copies): Use vect_get_num_vectors.
101         (vect_analyze_data_ref_dependences): Change max_vf from an int *
102         to an unsigned int *.
103         (vect_analyze_data_refs): Change min_vf from an int * to a
104         poly_uint64 *.
105         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
106         than an unsigned HOST_WIDE_INT.
107         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
108         (vect_analyze_data_ref_dependence): Change max_vf from an int *
109         to an unsigned int *.
110         (vect_analyze_data_ref_dependences): Likewise.
111         (vect_compute_data_ref_alignment): Handle polynomial vf.
112         (vect_enhance_data_refs_alignment): Likewise.
113         (vect_prune_runtime_alias_test_list): Likewise.
114         (vect_shift_permute_load_chain): Likewise.
115         (vect_supportable_dr_alignment): Likewise.
116         (dependence_distance_ge_vf): Take the vectorization factor as a
117         poly_uint64 rather than an unsigned HOST_WIDE_INT.
118         (vect_analyze_data_refs): Change min_vf from an int * to a
119         poly_uint64 *.
120         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
121         vfm1 as a poly_uint64 rather than an int.  Make the same change
122         for the returned bound_scalar.
123         (vect_gen_vector_loop_niters): Handle polynomial vf.
124         (vect_do_peeling): Likewise.  Update call to
125         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
126         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
127         be constant.
128         * tree-vect-loop.c (vect_determine_vectorization_factor)
129         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
130         (vect_get_known_peeling_cost): Likewise.
131         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
132         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
133         (vect_transform_loop): Likewise.  Use the lowest possible VF when
134         updating the upper bounds of the loop.
135         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
136         rather than an int.
137         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
138         polynomial unroll factors.
139         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
140         (vect_make_slp_decision): Likewise.
141         (vect_supported_load_permutation_p): Likewise, and polynomial
142         vf too.
143         (vect_analyze_slp_cost): Handle polynomial vf.
144         (vect_slp_analyze_node_operations): Likewise.
145         (vect_slp_analyze_bb_1): Likewise.
146         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
147         than an unsigned HOST_WIDE_INT.
148         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
149         (vectorizable_load): Handle polynomial vf.
150         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
151         a poly_uint64.
152         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
154 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
155             Alan Hayward  <alan.hayward@arm.com>
156             David Sherwood  <david.sherwood@arm.com>
158         * match.pd: Handle bit operations involving three constants
159         and try to fold one pair.
161 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
163         * tree-vect-loop-manip.c: Include gimple-fold.h.
164         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
165         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
166         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
167         Add a path that uses a step of VF instead of 1, but disable it
168         for now.
169         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
170         and niters_no_overflow parameters.  Update calls to
171         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
172         Create a new SSA name if the latter choses to use a ste other
173         than zero, and return it via niters_vector_mult_vf_var.
174         * tree-vect-loop.c (vect_transform_loop): Update calls to
175         vect_do_peeling, vect_gen_vector_loop_niters and
176         slpeel_make_loop_iterate_ntimes.
177         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
178         (vect_gen_vector_loop_niters): Update declarations after above changes.
180 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
182         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
183         128-bit round to integer instructions.
184         (ceil<mode>2): Likewise.
185         (btrunc<mode>2): Likewise.
186         (round<mode>2): Likewise.
188 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
190         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
191         unaligned VSX load/store on P8/P9.
192         (expand_block_clear): Allow the use of unaligned VSX
193         load/store on P8/P9.
195 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
197         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
198         New function.
199         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
200         swap associated with both a load and a store.
202 2018-01-02  Andrew Waterman  <andrew@sifive.com>
204         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
205         * config/riscv/riscv.md (clear_cache): Use it.
207 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
209         * web.c: Remove out-of-date comment.
211 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
213         * expr.c (fixup_args_size_notes): Check that any existing
214         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
215         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
216         (emit_single_push_insn): ...here.
218 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
220         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
221         (const_vector_encoded_nelts): New function.
222         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
223         (const_vector_int_elt, const_vector_elt): Declare.
224         * emit-rtl.c (const_vector_int_elt_1): New function.
225         (const_vector_elt): Likewise.
226         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
227         of CONST_VECTOR_ELT.
229 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
231         * expr.c: Include rtx-vector-builder.h.
232         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
233         directly on the tree encoding.
234         (const_vector_from_tree): Likewise.
235         * optabs.c: Include rtx-vector-builder.h.
236         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
237         sequence of "u" values.
238         * vec-perm-indices.c: Include rtx-vector-builder.h.
239         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
240         directly on the vec_perm_indices encoding.
242 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
244         * doc/rtl.texi (const_vector): Describe new encoding scheme.
245         * Makefile.in (OBJS): Add rtx-vector-builder.o.
246         * rtx-vector-builder.h: New file.
247         * rtx-vector-builder.c: Likewise.
248         * rtl.h (rtx_def::u2): Add a const_vector field.
249         (CONST_VECTOR_NPATTERNS): New macro.
250         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
251         (CONST_VECTOR_DUPLICATE_P): Likewise.
252         (CONST_VECTOR_STEPPED_P): Likewise.
253         (CONST_VECTOR_ENCODED_ELT): Likewise.
254         (const_vec_duplicate_p): Check for a duplicated vector encoding.
255         (unwrap_const_vec_duplicate): Likewise.
256         (const_vec_series_p): Check for a non-duplicated vector encoding.
257         Say that the function only returns true for integer vectors.
258         * emit-rtl.c: Include rtx-vector-builder.h.
259         (gen_const_vec_duplicate_1): Delete.
260         (gen_const_vector): Call gen_const_vec_duplicate instead of
261         gen_const_vec_duplicate_1.
262         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
263         (gen_const_vec_duplicate): Use rtx_vector_builder.
264         (gen_const_vec_series): Likewise.
265         (gen_rtx_CONST_VECTOR): Likewise.
266         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
267         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
268         Build a new vector rather than modifying a CONST_VECTOR in-place.
269         (handle_special_swappables): Update call accordingly.
270         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
271         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
272         Build a new vector rather than modifying a CONST_VECTOR in-place.
273         (handle_special_swappables): Update call accordingly.
275 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
277         * simplify-rtx.c (simplify_const_binary_operation): Use
278         CONST_VECTOR_ELT instead of XVECEXP.
280 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
282         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
283         the selector elements to be different from the data elements
284         if the selector is a VECTOR_CST.
285         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
286         ssizetype for the selector.
288 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
290         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
291         before testing each element individually.
292         * tree-vect-generic.c (lower_vec_perm): Likewise.
294 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
296         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
297         * selftest-run-tests.c (selftest::run_tests): Call it.
298         * vector-builder.h (vector_builder::operator ==): New function.
299         (vector_builder::operator !=): Likewise.
300         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
301         (vec_perm_indices::all_from_input_p): New function.
302         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
303         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
304         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
305         instead of reading the VECTOR_CST directly.  Detect whether both
306         vector inputs are the same before constructing the vec_perm_indices,
307         and update the number of inputs argument accordingly.  Use the
308         utility functions added above.  Only construct sel2 if we need to.
310 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
312         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
313         the broadcast of the low byte.
314         (expand_mult_highpart): Use an explicit encoding for the permutes.
315         * optabs-query.c (can_mult_highpart_p): Likewise.
316         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
317         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
318         (vectorizable_bswap): Likewise.
319         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
320         explicit encoding for the power-of-2 permutes.
321         (vect_permute_store_chain): Likewise.
322         (vect_grouped_load_supported): Likewise.
323         (vect_permute_load_chain): Likewise.
325 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
327         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
328         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
329         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
330         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
331         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
332         (vect_gen_perm_mask_any): Likewise.
334 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
336         * int-vector-builder.h: New file.
337         * vec-perm-indices.h: Include int-vector-builder.h.
338         (vec_perm_indices): Redefine as an int_vector_builder.
339         (auto_vec_perm_indices): Delete.
340         (vec_perm_builder): Redefine as a stand-alone class.
341         (vec_perm_indices::vec_perm_indices): New function.
342         (vec_perm_indices::clamp): Likewise.
343         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
344         (vec_perm_indices::new_vector): New function.
345         (vec_perm_indices::new_expanded_vector): Update for new
346         vec_perm_indices class.
347         (vec_perm_indices::rotate_inputs): New function.
348         (vec_perm_indices::all_in_range_p): Operate directly on the
349         encoded form, without computing elided elements.
350         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
351         encoding.  Update for new vec_perm_indices class.
352         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
353         the given vec_perm_builder.
354         (expand_vec_perm_var): Update vec_perm_builder constructor.
355         (expand_mult_highpart): Use vec_perm_builder instead of
356         auto_vec_perm_indices.
357         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
358         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
359         or double series encoding as appropriate.
360         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
361         vec_perm_indices instead of auto_vec_perm_indices.
362         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
363         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
364         (vect_permute_store_chain): Likewise.
365         (vect_grouped_load_supported): Likewise.
366         (vect_permute_load_chain): Likewise.
367         (vect_shift_permute_load_chain): Likewise.
368         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
369         (vect_transform_slp_perm_load): Likewise.
370         (vect_schedule_slp_instance): Likewise.
371         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
372         (vectorizable_mask_load_store): Likewise.
373         (vectorizable_bswap): Likewise.
374         (vectorizable_store): Likewise.
375         (vectorizable_load): Likewise.
376         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
377         vec_perm_indices instead of auto_vec_perm_indices.  Use
378         tree_to_vec_perm_builder to read the vector from a tree.
379         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
380         vec_perm_builder instead of a vec_perm_indices.
381         (have_whole_vector_shift): Use vec_perm_builder and
382         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
383         truncation to calc_vec_perm_mask_for_shift.
384         (vect_create_epilog_for_reduction): Likewise.
385         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
386         from auto_vec_perm_indices to vec_perm_indices.
387         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
388         instead of changing individual elements.
389         (aarch64_vectorize_vec_perm_const): Use new_vector to install
390         the vector in d.perm.
391         * config/arm/arm.c (expand_vec_perm_d::perm): Change
392         from auto_vec_perm_indices to vec_perm_indices.
393         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
394         instead of changing individual elements.
395         (arm_vectorize_vec_perm_const): Use new_vector to install
396         the vector in d.perm.
397         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
398         Update vec_perm_builder constructor.
399         (rs6000_expand_interleave): Likewise.
400         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
401         (rs6000_expand_interleave): Likewise.
403 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
405         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
406         to qimode could truncate the indices.
407         * optabs.c (expand_vec_perm_var): Likewise.
409 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
411         * Makefile.in (OBJS): Add vec-perm-indices.o.
412         * vec-perm-indices.h: New file.
413         * vec-perm-indices.c: Likewise.
414         * target.h (vec_perm_indices): Replace with a forward class
415         declaration.
416         (auto_vec_perm_indices): Move to vec-perm-indices.h.
417         * optabs.h: Include vec-perm-indices.h.
418         (expand_vec_perm): Delete.
419         (selector_fits_mode_p, expand_vec_perm_var): Declare.
420         (expand_vec_perm_const): Declare.
421         * target.def (vec_perm_const_ok): Replace with...
422         (vec_perm_const): ...this new hook.
423         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
424         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
425         * doc/tm.texi: Regenerate.
426         * optabs.def (vec_perm_const): Delete.
427         * doc/md.texi (vec_perm_const): Likewise.
428         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
429         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
430         expand_vec_perm for constant permutation vectors.  Assert that
431         the mode of variable permutation vectors is the integer equivalent
432         of the mode that is being permuted.
433         * optabs-query.h (selector_fits_mode_p): Declare.
434         * optabs-query.c: Include vec-perm-indices.h.
435         (selector_fits_mode_p): New function.
436         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
437         is defined, instead of checking whether the vec_perm_const_optab
438         exists.  Use targetm.vectorize.vec_perm_const instead of
439         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
440         fit in the vector mode before using a variable permute.
441         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
442         vec_perm_indices instead of an rtx.
443         (expand_vec_perm): Replace with...
444         (expand_vec_perm_const): ...this new function.  Take the selector
445         as a vec_perm_indices rather than an rtx.  Also take the mode of
446         the selector.  Update call to shift_amt_for_vec_perm_mask.
447         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
448         Use vec_perm_indices::new_expanded_vector to expand the original
449         selector into bytes.  Check whether the indices fit in the vector
450         mode before using a variable permute.
451         (expand_vec_perm_var): Make global.
452         (expand_mult_highpart): Use expand_vec_perm_const.
453         * fold-const.c: Includes vec-perm-indices.h.
454         * tree-ssa-forwprop.c: Likewise.
455         * tree-vect-data-refs.c: Likewise.
456         * tree-vect-generic.c: Likewise.
457         * tree-vect-loop.c: Likewise.
458         * tree-vect-slp.c: Likewise.
459         * tree-vect-stmts.c: Likewise.
460         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
461         Delete.
462         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
463         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
464         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
465         (aarch64_vectorize_vec_perm_const): ...this new function.
466         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
467         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
468         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
469         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
470         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
471         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
472         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
473         into...
474         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
475         check for NEON modes.
476         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
477         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
478         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
479         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
480         into...
481         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
482         the old VEC_PERM_CONST conditions.
483         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
484         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
485         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
486         (ia64_vectorize_vec_perm_const_ok): Merge into...
487         (ia64_vectorize_vec_perm_const): ...this new function.
488         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
489         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
490         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
491         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
492         * config/mips/mips.c (mips_expand_vec_perm_const)
493         (mips_vectorize_vec_perm_const_ok): Merge into...
494         (mips_vectorize_vec_perm_const): ...this new function.
495         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
496         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
497         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
498         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
499         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
500         (rs6000_expand_vec_perm_const): Delete.
501         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
502         Delete.
503         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
504         (altivec_expand_vec_perm_const_le): Take each operand individually.
505         Operate on constant selectors rather than rtxes.
506         (altivec_expand_vec_perm_const): Likewise.  Update call to
507         altivec_expand_vec_perm_const_le.
508         (rs6000_expand_vec_perm_const): Delete.
509         (rs6000_vectorize_vec_perm_const_ok): Delete.
510         (rs6000_vectorize_vec_perm_const): New function.
511         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
512         an element count and rtx array.
513         (rs6000_expand_extract_even): Update call accordingly.
514         (rs6000_expand_interleave): Likewise.
515         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
516         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
517         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
518         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
519         (rs6000_expand_vec_perm_const): Delete.
520         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
521         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
522         (altivec_expand_vec_perm_const_le): Take each operand individually.
523         Operate on constant selectors rather than rtxes.
524         (altivec_expand_vec_perm_const): Likewise.  Update call to
525         altivec_expand_vec_perm_const_le.
526         (rs6000_expand_vec_perm_const): Delete.
527         (rs6000_vectorize_vec_perm_const_ok): Delete.
528         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
529         reference to the SPE evmerge intructions.
530         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
531         an element count and rtx array.
532         (rs6000_expand_extract_even): Update call accordingly.
533         (rs6000_expand_interleave): Likewise.
534         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
535         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
536         new function.
537         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
539 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
541         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
542         vector mode and that that mode matches the mode of the data
543         being permuted.
544         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
545         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
546         directly using expand_vec_perm_1 when forcing selectors into
547         registers.
548         (expand_vec_perm_var): New function, split out from expand_vec_perm.
550 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
552         * optabs-query.h (can_vec_perm_p): Delete.
553         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
554         * optabs-query.c (can_vec_perm_p): Split into...
555         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
556         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
557         particular selector is valid.
558         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
559         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
560         (vect_grouped_load_supported): Likewise.
561         (vect_shift_permute_load_chain): Likewise.
562         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
563         (vect_transform_slp_perm_load): Likewise.
564         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
565         (vectorizable_bswap): Likewise.
566         (vect_gen_perm_mask_checked): Likewise.
567         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
568         implementations of variable permutation vectors into account
569         when deciding which selector to use.
570         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
571         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
572         with a false third argument.
573         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
574         to test whether the constant selector is valid and can_vec_perm_var_p
575         to test whether a variable selector is valid.
577 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
579         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
580         * optabs-query.c (can_vec_perm_p): Likewise.
581         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
582         instead of vec_perm_indices.
583         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
584         (vect_gen_perm_mask_checked): Likewise,
585         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
586         (vect_gen_perm_mask_checked): Likewise,
588 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
590         * optabs-query.h (qimode_for_vec_perm): Declare.
591         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
592         (qimode_for_vec_perm): ...this new function.
593         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
595 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
597         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
598         does not have a conditional at the top.
600 2018-01-02  Richard Biener  <rguenther@suse.de>
602         * ipa-inline.c (big_speedup_p): Fix expression.
604 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
606         PR target/81616
607         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
608         for generic 4->6.
610 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
612         PR target/81616
613         Generic tuning.
614         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
615         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
616         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
617         cond_taken_branch_cost 3->4.
619 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
621         PR tree-optimization/83581
622         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
623         TODO_cleanup_cfg if any changes have been made.
625         PR middle-end/83608
626         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
627         convert_modes if target mode has the right side, but different mode
628         class.
630         PR middle-end/83609
631         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
632         last argument when extracting from CONCAT.  If either from_real or
633         from_imag is NULL, use expansion through memory.  If result is not
634         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
635         the parts directly to inner mode, if even that fails, use expansion
636         through memory.
638         PR middle-end/83623
639         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
640         check for bswap in mode rather than HImode and use that in expand_unop
641         too.
643 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
645         PR target/83536
646         * config/i386/i386.c (ix86_attribute_table): Remove excess
647         initializer for "shared" attribute.
649 2017-12-30  Tom de Vries  <tom@codesourcery.com>
651         PR libgomp/83046
652         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
653         DECL_PRESERVE_P.
654         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
655         offload_funcs entries that no longer have a corresponding cgraph_node.
656         Mark the remaining ones as DECL_PRESERVE_P.
657         (output_lto): Call prune_offload_funcs.
659 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
661         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
662         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
663         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
664         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
665         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
666         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
667         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
668         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
669         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
670         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
671         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
672         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
673         vaesenc_<mode>, vpclmulqdq_<mode>,
674         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
676 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
678         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
679         _Float<N> and _Float<N>X variants for rounding built-in
680         functions.
681         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
682         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
683         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
684         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
685         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
686         * builtins.c (mathfn_built_in_2): Likewise.
687         * internal-fn.def (CEIL): Likewise.
688         (FLOOR): Likewise.
689         (NEARBYINT): Likewise.
690         (RINT): Likewise.
691         (ROUND): Likewise.
692         (TRUNC): Likewise.
693         * convert.c (convert_to_integer_1): Likewise.
694         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
695         (integer_valued_real_call_p): Likewise.
696         * fold-const-call.c (fold_const_call_ss): Likewise.
697         * gencfn-macros.c (print_case_cfn): Change CFN and operator
698         printers to take a const char * suffix instead of a bool.
699         (print_define_operator_list): Likewise.
700         (fltall_suffixes): New list of suffixes, that include the
701         traditional suffixes as well as all of the _Float<N> and
702         _Float<N>X suffixes.
703         (main): For _Float<N> and _Float<N>X functions, emit both
704         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
705         has the _Float<N> and _Float<N>X case names or operators.  The
706         <name>_ALL variant has both the traditional and the
707         _Float<N>/_Float<N>X case names or operators.
708         * match.pd (COPYSIGN optimizations): Provide optimizations for
709         _Float<N> and _Float<N>X types where possible.
710         (MIN/MAX optimizations): Likewise.
711         (sqrt optimizations): Likewise.
712         (rounding optimizations): Likewise.
714 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
716         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
717         gen_int_mode rather than GEN_INT.
718         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
720 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
722         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
723         instead of CONSTANT_P.
724         (gen_vec_series): Likewise.
725         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
727 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
729         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
730         CC_STATUS_INIT.
732 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
734         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
735         (aarch64_ld1x2<VDC:mode>): Likewise.
736         (aarch64_simd_ld1<mode>_x2): Likewise.
737         (aarch64_simd_ld1<mode>_x2): Likewise.
738         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
739         (vld1_s8_x2): Likewise.
740         (vld1_u16_x2): Likewise.
741         (vld1_s16_x2): Likewise.
742         (vld1_u32_x2): Likewise.
743         (vld1_s32_x2): Likewise.
744         (vld1_u64_x2): Likewise.
745         (vld1_s64_x2): Likewise.
746         (vld1_f16_x2): Likewise.
747         (vld1_f32_x2): Likewise.
748         (vld1_f64_x2): Likewise.
749         (vld1_p8_x2): Likewise.
750         (vld1_p16_x2): Likewise.
751         (vld1_p64_x2): Likewise.
752         (vld1q_u8_x2): Likewise.
753         (vld1q_s8_x2): Likewise.
754         (vld1q_u16_x2): Likewise.
755         (vld1q_s16_x2): Likewise.
756         (vld1q_u32_x2): Likewise.
757         (vld1q_s32_x2): Likewise.
758         (vld1q_u64_x2): Likewise.
759         (vld1q_s64_x2): Likewise.
760         (vld1q_f16_x2): Likewise.
761         (vld1q_f32_x2): Likewise.
762         (vld1q_f64_x2): Likewise.
763         (vld1q_p8_x2): Likewise.
764         (vld1q_p16_x2): Likewise.
765         (vld1q_p64_x2): Likewise.
767 2017-12-27  Martin Liska  <mliska@suse.cz>
769         PR tree-optimization/83552
770         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
771         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
773 2017-12-27  Tom de Vries  <tom@codesourcery.com>
775         * config/nvptx/nvptx.c (nvptx_option_override): Disable
776         -gstatement-frontiers.
778 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
780         PR rtl-optimization/83513
781         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
782         before priority comparison.
784 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
786         PR target/83488
787         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
788         ix86_isa_flags2 to ix86_isa_flags.
789         * config/i386/i386-c.c (ix86_target_macros_internal): Test
790         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
791         isa_flags rather than isa_flags2.
792         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
793         and -mavx512bitalg from isa2_opts to isa_opts.
794         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
795         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
796         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
797         Formatting fixes.
798         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
799         OPTION_MASK_ISA_AVX512F ored with another option similarly to
800         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
801         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
802         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
803         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
804         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
805         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
806         builtins from bdesc_args2 to bdesc_args section.
807         (__builtin_ia32_compressstoreuqi512_mask,
808         __builtin_ia32_compressstoreuhi512_mask,
809         __builtin_ia32_compressstoreuqi256_mask,
810         __builtin_ia32_expandloadqi512_mask,
811         __builtin_ia32_expandloadqi512_maskz,
812         __builtin_ia32_expandloadhi512_mask,
813         __builtin_ia32_expandloadhi512_maskz,
814         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
815         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
816         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
817         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
818         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
819         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
820         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
821         __builtin_ia32_vpshldv_v32hi_maskz,
822         __builtin_ia32_vpopcountb_v64qi_mask,
823         __builtin_ia32_vpopcountw_v32hi_mask,
824         __builtin_ia32_vpshufbitqmb512_mask,
825         __builtin_ia32_vpshufbitqmb256_mask): Add
826         " | OPTION_MASK_ISA_AVX512BW".
827         (__builtin_ia32_expandloadqi256_mask,
828         __builtin_ia32_expandloadqi256_maskz,
829         __builtin_ia32_vpopcountb_v32qi_mask): Add
830         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
831         (__builtin_ia32_expandloadhi256_mask,
832         __builtin_ia32_expandloadhi256_maskz,
833         __builtin_ia32_expandloadqi128_mask,
834         __builtin_ia32_expandloadqi128_maskz,
835         __builtin_ia32_expandloadhi128_mask,
836         __builtin_ia32_expandloadhi128_maskz,
837         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
838         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
839         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
840         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
841         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
842         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
843         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
844         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
845         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
846         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
847         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
848         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
849         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
850         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
851         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
852         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
853         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
854         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
855         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
856         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
857         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
858         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
859         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
860         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
861         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
862         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
863         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
864         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
865         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
866         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
867         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
868         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
869         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
870         __builtin_ia32_vpopcountw_v8hi_mask): Add
871         " | OPTION_MASK_ISA_AVX512VL".
872         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
873         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
874         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
875         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
876         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
877         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
878         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
879         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
880         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
881         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
882         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
883         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
884         _mm_mask_bitshuffle_epi64_mask): Likewise.
885         * common/config/i386/i386-common.c
886         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
887         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
888         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
889         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
890         OPTION_MASK_ISA_AVX512BITALG_UNSET.
891         (OPTION_MASK_ISA2_AVX512F_UNSET,
892         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
893         (ix86_handle_option): For -mno-general-regs-only, clear from
894         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
895         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
896         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
897         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
898         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
899         for moving from ix86_isa_flags2 to ix86_isa_flags.
901 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
903         PR c++/83553
904         * fold-const.c (struct contains_label_data): New type.
905         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
906         inside of a SWITCH_BODY seen during the walk.
907         (contains_label_p): Use walk_tree instead of
908         walk_tree_without_duplicates, prepare data for contains_label_1 and
909         provide own pset.
911 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
913         PR debug/83547
914         * tree-iterator.c (alloc_stmt_list): Start with cleared
915         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
916         or old one reused.
918         PR target/83488
919         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
920         enable avx512f explicitly in #pragma GCC target.
921         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
922         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
923         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
924         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
925         __builtin_ia32_vpdpbusds_v8si_mask,
926         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
927         __builtin_ia32_vpdpbusds_v4si_mask,
928         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
929         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
930         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
931         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
932         __builtin_ia32_vpdpwssds_v8si_mask,
933         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
934         __builtin_ia32_vpdpwssds_v4si_mask,
935         __builtin_ia32_vpdpwssds_v4si_maskz): Use
936         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
937         just OPTION_MASK_ISA_AVX512VNNI.
939 2017-12-22  Martin Jambor  <mjambor@suse.cz>
941         PR lto/82027
942         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
943         clones.
945 2017-12-22  Julia Koval  <julia.koval@intel.com>
946             Sebastian Peryt  <sebastian.peryt@intel.com>
948         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
949         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
950         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
951         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
952         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
953         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
954         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
955         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
956         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
957         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
958         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
959         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
960         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
961         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
962         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
963         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
964         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
965         _mm256_maskz_popcnt_epi64): New intrinsics.
966         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
967         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
968         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
969         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
970         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
971         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
972         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
973         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
974         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
975         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
976         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
977         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
978         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
979         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
980         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
981         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
982         __builtin_ia32_vpshufbitqmb256_mask,
983         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
984         * config/i386/i386-c.c (__AVX512BITALG__): New.
985         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
986         (ix86_valid_target_attribute_inner_p): Ditto.
987         (ix86_expand_args_builtin): Handle new types.
988         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
989         * config/i386/i386.opt: Add -mavx512bitalg.
990         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
991         avx512bitalgintrin.h.
992         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
993         (vpopcount<mode><mask_name>): Add more types.
994         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
995         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
997 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
999         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1000         Or in OPTION_MASK_ISA_AVX512F_SET.
1001         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1002         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1003         (ix86_handle_option): Adjust for
1004         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1005         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1006         section to ARGS.
1007         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1008         isa_flag instead of isa_flag2.
1009         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1010         isa_opts2 to isa_opts.
1011         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1012         to ix86_isa_flags.
1014 2017-12-22  Mike Stump  <mikestump@comcast.net>
1015             Eric Botcazou  <ebotcazou@adacore.com>
1017         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1019 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1021         PR middle-end/83487
1022         * config/i386/i386.c (ix86_function_arg_boundary): Return
1023         PARM_BOUNDARY for TYPE_EMPTY_P types.
1025         PR c/83448
1026         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1027         if navail is >= dir.len.
1029 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1031         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1032         triplet for ilp32.
1034 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1036         PR rtl-optimization/80747
1037         PR rtl-optimization/83512
1038         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1039         succ edge from ENTRY, copy partition from e->dest to the newly
1040         created bb.
1041         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1042         ENTRY, use BB_PARTITION of its successor block as current_partition.
1043         Don't copy partition when splitting succ edge from ENTRY.
1045         PR tree-optimization/83523
1046         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1047         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1048         (convert_mult_to_fma): Likewise.
1050         PR tree-optimization/83521
1051         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1052         gimple_build_assign without code on result of
1053         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1054         a VIEW_CONVERT_EXPR.
1056 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1057             Steve Ellcey  <sellcey@cavium.com>
1059         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1060         multi-arch for ilp32.
1062 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1064         PR target/83467
1065         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1066         constraints to operand 2.
1067         (*ashl<mode>3_mask_1): Ditto.
1068         (*<shift_insn><mode>3_mask): Ditto.
1069         (*<shift_insn><mode>3_mask_1): Ditto.
1070         (*<rotate_insn><mode>3_mask): Ditto.
1071         (*<rotate_insn><mode>3_mask_1): Ditto.
1073 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1075         * reorg.c (make_return_insns): Reemit each insn with its own location.
1077 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1079         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1080         generation for cases where splatting a value is not useful.
1081         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1082         across a vec_duplicate and a paradoxical subreg forming a vector
1083         mode to a vec_concat.
1085 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1087         * combine.c (simplify_set): Do not transform subregs to zero_extends
1088         if the destination is not a scalar int mode.
1090 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1092         PR c++/82872
1093         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1094         shared zero if the input has overflowed.
1096 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1098         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1099         support for Cortex-A55 and Cortex-A75.
1101 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1103         * common/config/arm/arm-common.c (compare_opt_names): Add function
1104         comment.  Use strcmp instead of manual loop.
1106 2017-12-21  Martin Liska  <mliska@suse.cz>
1108         PR gcov-profile/83509
1109         * gcov-dump.c (dump_gcov_file): Do not read info about
1110         support_unexecuted_blocks for gcda files.
1112 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1114         PR rtl-optimization/82973
1115         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1116         (valid_for_const_vector_p): ... this.
1117         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1118         (valid_for_const_vector_p): ... this.  Adjust function comment.
1119         (gen_vec_duplicate): Adjust caller.
1120         * optabs.c (expand_vector_broadcast): Likewise.
1121         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1122         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1123         constant.
1124         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1125         instead of GET_CODE == CONST_FIXED.
1126         (simplify_subreg): Use CONST_FIXED_P macro instead of
1127         GET_CODE == CONST_FIXED.
1129         PR target/83488
1130         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1131         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1132         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1133         (ix86_option_override_internal): Adjust for
1134         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1135         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1136         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1137         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1138         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1139         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1140         def_builtin2 for CET builtins.
1141         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1142         up formatting in IX86_BUILTIN_RDPID code.
1143         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1144         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1145         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1146         ix86_isa_flags2 to ix86_isa_flags.
1147         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1148         ix86_isa_flags2.
1149         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1150         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1151         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1152         of isa_flag2.
1153         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1154         Or in OPTION_MASK_ISA_AVX512F_SET.
1155         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1156         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1157         (ix86_handle_option): Adjust for
1158         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1159         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1161 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1162             Alan Hayward  <alan.hayward@arm.com>
1163             David Sherwood  <david.sherwood@arm.com>
1165         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1166         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1167         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1168         Track polynomial offsets.
1170 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1171             Alan Hayward  <alan.hayward@arm.com>
1172             David Sherwood  <david.sherwood@arm.com>
1174         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1175         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1176         vector_alignment isn't known.
1178 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1179             Alan Hayward  <alan.hayward@arm.com>
1180             David Sherwood  <david.sherwood@arm.com>
1182         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1183         field.
1184         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1185         (vect_loop_versioning): Take the loop versioning threshold as a
1186         separate parameter.
1187         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1188         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1189         versioning_threshold.
1190         (vect_analyze_loop_2): Compute the loop versioning threshold
1191         whenever loop versioning is needed, and store it in the new
1192         field rather than combining it with the cost model threshold.
1193         (vect_transform_loop): Update call to vect_loop_versioning.
1194         Try to combine the loop versioning and cost thresholds here.
1196 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1197             Alan Hayward  <alan.hayward@arm.com>
1198             David Sherwood  <david.sherwood@arm.com>
1200         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1201         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1202         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1203         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1204         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1205         (group_compare_offset): Update accordingly.
1206         (split_small_address_groups_p): Likewise.
1207         (record_use): Take addr_offset as a poly_uint64 rather than
1208         an unsigned HOST_WIDE_INT.
1209         (strip_offset): Return the offset as a poly_uint64 rather than
1210         an unsigned HOST_WIDE_INT.
1211         (record_group_use, split_address_groups): Track polynomial offsets.
1212         (add_iv_candidate_for_use): Likewise.
1213         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1214         than a HOST_WIDE_INT.
1215         (strip_offset_1): Return the offset as a poly_int64 rather than
1216         a HOST_WIDE_INT.
1218 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1219             Alan Hayward  <alan.hayward@arm.com>
1220             David Sherwood  <david.sherwood@arm.com>
1222         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1223         rather than a HOST_WIDE_INT.
1224         * tree.c (get_binfo_at_offset): Likewise.
1226 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1227             Alan Hayward  <alan.hayward@arm.com>
1228             David Sherwood  <david.sherwood@arm.com>
1230         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1231         rather than a HOST_WIDE_INT.
1232         * tree-sra.c (build_ref_for_offset): Likewise.
1234 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1235             Alan Hayward  <alan.hayward@arm.com>
1236             David Sherwood  <david.sherwood@arm.com>
1238         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1239         than an offset_int.
1240         * tree.c (mem_ref_offset): Likewise.
1241         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1242         * builtins.c (get_object_alignment_2): Likewise.
1243         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1244         * gimple-fold.c (get_base_constructor): Likewise.
1245         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1246         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1247         Likewise.
1248         * ipa-polymorphic-call.c
1249         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1250         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1251         (get_ancestor_addr_info): Likewise.
1252         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1253         * match.pd: Likewise.
1254         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1255         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1256         * tree-eh.c (tree_could_trap_p): Likewise.
1257         * tree-object-size.c (addr_object_size): Likewise.
1258         * tree-ssa-address.c (copy_ref_info): Likewise.
1259         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1260         (indirect_refs_may_alias_p): Likewise.
1261         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1262         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1263         (non_rewritable_mem_ref_base): Likewise.
1264         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1265         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1266         * varasm.c (decode_addr_const): Likewise.
1268 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1269             Alan Hayward  <alan.hayward@arm.com>
1270             David Sherwood  <david.sherwood@arm.com>
1272         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1273         offsets for MEM_REFs.
1275 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1276             Alan Hayward  <alan.hayward@arm.com>
1277             David Sherwood  <david.sherwood@arm.com>
1279         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1280         as a poly_uint64 rather than an unsigned int.
1281         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1283 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1284             Alan Hayward  <alan.hayward@arm.com>
1285             David Sherwood  <david.sherwood@arm.com>
1287         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1288         to poly_int64.
1289         (decode_addr_const): Update accordingly.
1291 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1292             Alan Hayward  <alan.hayward@arm.com>
1293             David Sherwood  <david.sherwood@arm.com>
1295         * tree.h (bit_field_size, bit_field_offset): New functions.
1296         * hsa-gen.c (gen_hsa_addr): Use them.
1297         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1298         (simplify_vector_constructor): Likewise.
1299         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1300         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1301         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1302         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1303         tree_fits_uhwi_p.
1305 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1306             Alan Hayward  <alan.hayward@arm.com>
1307             David Sherwood  <david.sherwood@arm.com>
1309         * expr.h (emit_group_load, emit_group_load_into_temps)
1310         (emit_group_store): Take the size as a poly_int64 rather than an int.
1311         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1312         (emit_group_load_into_temp, emit_group_store): Likewise.
1314 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1315             Alan Hayward  <alan.hayward@arm.com>
1316             David Sherwood  <david.sherwood@arm.com>
1318         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1319         an unsigned int to a poly_uint64.
1320         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1321         sizes as poly_uint64s rather than unsigned ints.
1322         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1323         Likewise.
1325 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1326             Alan Hayward  <alan.hayward@arm.com>
1327             David Sherwood  <david.sherwood@arm.com>
1329         * lra-constraints.c (emit_inc): Change inc_amount from an int
1330         to a poly_int64.
1332 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1333             Alan Hayward  <alan.hayward@arm.com>
1334             David Sherwood  <david.sherwood@arm.com>
1336         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1337         to a poly_uint64.
1338         (add_stack_var, stack_var_cmp, partition_stack_vars)
1339         (dump_stack_var_partition): Update accordingly.
1340         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1341         than a HOST_WIDE_INT.
1342         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1343         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1344         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1345         rather than a HOST_WIDE_INT.
1347 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1348             Alan Hayward  <alan.hayward@arm.com>
1349             David Sherwood  <david.sherwood@arm.com>
1351         * target.def (return_pops_args): Treat both the input and output
1352         sizes as poly_int64s rather than HOST_WIDE_INTS.
1353         * targhooks.h (default_return_pops_args): Update accordingly.
1354         * targhooks.c (default_return_pops_args): Likewise.
1355         * doc/tm.texi: Regenerate.
1356         * emit-rtl.h (incoming_args): Change pops_args, size and
1357         outgoing_args_size from int to poly_int64_pod.
1358         * function.h (expr_status): Change x_pending_stack_adjust and
1359         x_stack_pointer_delta from int to poly_int64.
1360         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1361         (ARGS_SIZE_RTX): Update accordingly.
1362         * calls.c (highest_outgoing_arg_in_use): Change from int to
1363         unsigned int.
1364         (stack_usage_watermark, stored_args_watermark): New variables.
1365         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1366         (emit_call_1): Change the stack_size and rounded_stack_size
1367         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1368         as a poly_int64.
1369         (save_fixed_argument_area): Check stack_usage_watermark.
1370         (initialize_argument_information): Change old_pending_adj from
1371         a HOST_WIDE_INT * to a poly_int64_pod *.
1372         (compute_argument_block_size): Return the size as a poly_int64
1373         rather than an int.
1374         (finalize_must_preallocate): Track polynomial argument sizes.
1375         (compute_argument_addresses): Likewise.
1376         (internal_arg_pointer_based_exp): Track polynomial offsets.
1377         (mem_overlaps_already_clobbered_arg_p): Rename to...
1378         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1379         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1380         Check stored_args_used_watermark.
1381         (load_register_parameters): Update accordingly.
1382         (check_sibcall_argument_overlap_1): Likewise.
1383         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1384         args size as a poly_int64 rather than an int.  Return a bool
1385         indicating whether the optimization was possible and return
1386         the new adjustment by reference.
1387         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1388         Update stored_args_watermark.
1389         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1390         (expand_call): Likewise.  Maintain stack_usage_watermark and
1391         stored_args_watermark.  Update calls to
1392         combine_pending_stack_adjustment_and_call.
1393         (emit_library_call_value_1): Handle polynomial argument sizes.
1394         Call stack_region_maybe_used_p and mark_stack_region_used.
1395         Maintain stack_usage_watermark.
1396         (store_one_arg): Likewise.  Update call to
1397         mem_overlaps_already_clobbered_arg_p.
1398         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1399         HOST_WIDE_INT.
1400         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1401         * config/microblaze/microblaze.c (microblaze_function_prologue):
1402         Likewise.
1403         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1404         TARGET_RETURN_POPS_ARGS interface.
1405         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1406         to HOST_WIDE_INT.
1407         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1408         * config/i386/i386.c (ix86_return_pops_args): Update for new
1409         TARGET_RETURN_POPS_ARGS interface.
1410         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1411         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1412         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1413         TARGET_RETURN_POPS_ARGS interface.
1414         * config/vax/vax.c (vax_return_pops_args): Likewise.
1415         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1416         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1417         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1418         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1419         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1420         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1421         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1422         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1423         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1424         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1425         and x_stack_pointer_delta from int to poly_int64.
1426         * dojump.c (do_pending_stack_adjust): Update accordingly.
1427         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1428         stack_pointer_deltas.
1429         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1430         (pad_to_arg_alignment): Track polynomial offsets.
1431         (assign_parm_find_stack_rtl): Likewise.
1432         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1433         * toplev.c (output_stack_usage): Update reference to
1434         current_function_pushed_stack_size.
1436 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1437             Alan Hayward  <alan.hayward@arm.com>
1438             David Sherwood  <david.sherwood@arm.com>
1440         * function.c (in_arg_offset, var_offset, dynamic_offset)
1441         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1442         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1443         rather than a HOST_WIDE_INT.
1444         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1445         (instantiate_virtual_regs_in_insn): Likewise.
1447 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1448             Alan Hayward  <alan.hayward@arm.com>
1449             David Sherwood  <david.sherwood@arm.com>
1451         * rtl.h (get_args_size, add_args_size_note): New functions.
1452         (find_args_size_adjust): Return a poly_int64 rather than a
1453         HOST_WIDE_INT.
1454         (fixup_args_size_notes): Likewise.  Make the same change to the
1455         end_args_size parameter.
1456         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1457         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1458         * calls.c (emit_call_1): Likewise.
1459         * explow.c (adjust_stack_1): Likewise.
1460         * cfgcleanup.c (old_insns_match_p): Update use of
1461         find_args_size_adjust.
1462         * combine.c (distribute_notes): Track polynomial arg sizes.
1463         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1464         end_true_args_size, beg_delay_args_size and end_delay_args_size
1465         from HOST_WIDE_INT to poly_int64.
1466         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1467         than a HOST_WIDE_INT.
1468         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1469         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1470         polynomial arg sizes.
1471         * emit-rtl.c (try_split): Use get_args_size.
1472         * recog.c (peep2_attempt): Likewise.
1473         * reload1.c (reload_as_needed): Likewise.
1474         * expr.c (find_args_size_adjust): Return the adjustment as a
1475         poly_int64 rather than a HOST_WIDE_INT.
1476         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1477         to a poly_int64 and change the return type in the same way.
1478         (emit_single_push_insn): Track polynomial arg sizes.
1480 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1481             Alan Hayward  <alan.hayward@arm.com>
1482             David Sherwood  <david.sherwood@arm.com>
1484         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1485         from HOST_WIDE_INT to poly_int64.
1486         * expr.c (push_block, emit_push_insn): Likewise.
1488 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1489             Alan Hayward  <alan.hayward@arm.com>
1490             David Sherwood  <david.sherwood@arm.com>
1492         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1493         to poly_int64.
1494         (get_frame_size): Return the size as a poly_int64 rather than a
1495         HOST_WIDE_INT.
1496         (frame_offset_overflow): Take the offset as a poly_int64 rather
1497         than a HOST_WIDE_INT.
1498         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1499         (assign_stack_temp): Likewise for the size.
1500         * function.c (get_frame_size): Return a poly_int64 rather than
1501         a HOST_WIDE_INT.
1502         (frame_offset_overflow): Take the offset as a poly_int64 rather
1503         than a HOST_WIDE_INT.
1504         (try_fit_stack_local): Take the start, length and size as poly_int64s
1505         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1506         rather than a HOST_WIDE_INT.
1507         (add_frame_space): Take the start and end as poly_int64s rather than
1508         HOST_WIDE_INTs.
1509         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1510         (assign_stack_temp): Likewise for the size.
1511         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1512         to poly_int64.
1513         (find_temp_slot_from_address): Handle polynomial offsets.
1514         (combine_temp_slots): Likewise.
1515         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1516         to poly_int64.
1517         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1518         poly_int64 rather than a HOST_WIDE_INT.
1519         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1520         than a HOST_WIDE_INT.
1521         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1522         polynomial frame offsets.
1523         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1524         as a poly_int64 rather than an int.
1525         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1526         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1527         * config/v850/v850.c (compute_frame_size): Likewise.
1528         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1529         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1530         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1531         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1532         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1533         rather than a HOST_WIDE_INT.
1534         * explow.c (get_dynamic_stack_base): Likewise.
1535         * final.c (final_start_function): Use the constant lower bound
1536         of the frame size for -Wframe-larger-than.
1537         * ira.c (do_reload): Adjust for new get_frame_size return type.
1538         * lra.c (lra): Likewise.
1539         * reload1.c (reload): Likewise.
1540         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1541         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1542         * rtlanal.c (get_initial_register_offset): Return the offset as
1543         a poly_int64 rather than a HOST_WIDE_INT.
1545 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1546             Alan Hayward  <alan.hayward@arm.com>
1547             David Sherwood  <david.sherwood@arm.com>
1549         * reload1.c (elim_table): Change initial_offset, offset and
1550         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1551         (offsets_at): Change the target array's element type from
1552         HOST_WIDE_INT to poly_int64_pod.
1553         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1554         (elimination_costs_in_insn, update_eliminable_offsets)
1555         (verify_initial_elim_offsets, set_offsets_for_label)
1556         (init_eliminable_invariants): Update after above changes.
1558 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1559             Alan Hayward  <alan.hayward@arm.com>
1560             David Sherwood  <david.sherwood@arm.com>
1562         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1563         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1564         (decomposition): Change start and end from HOST_WIDE_INT
1565         to poly_int64_pod.
1566         (decompose, immune_p): Update accordingly.
1567         (find_inc_amount): Return a poly_int64 rather than an int.
1568         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1569         rather than an int.
1571 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1572             Alan Hayward  <alan.hayward@arm.com>
1573             David Sherwood  <david.sherwood@arm.com>
1575         * tree.h (get_inner_reference): Return the bitsize and bitpos
1576         as poly_int64_pods rather than HOST_WIDE_INT.
1577         * fold-const.h (ptr_difference_const): Return the pointer difference
1578         as a poly_int64_pod rather than a HOST_WIDE_INT.
1579         * expr.c (get_inner_reference): Return the bitsize and bitpos
1580         as poly_int64_pods rather than HOST_WIDE_INT.
1581         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1582         offsets and sizes.
1583         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1584         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1585         (optimize_bit_field_compare): Update call to get_inner_reference.
1586         (decode_field_reference): Likewise.
1587         (fold_unary_loc): Track polynomial offsets and sizes.
1588         (split_address_to_core_and_offset): Return the bitpos as a
1589         poly_int64_pod rather than a HOST_WIDE_INT.
1590         (ptr_difference_const): Likewise for the pointer difference.
1591         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1592         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1593         * dbxout.c (dbxout_expand_expr): Likewise.
1594         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1595         (loc_list_from_tree_1, fortran_common): Likewise.
1596         * gimple-laddress.c (pass_laddress::execute): Likewise.
1597         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1598         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1599         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1600         * tree-affine.c (tree_to_aff_combination): Likewise.
1601         (get_inner_reference_aff): Likewise.
1602         * tree-data-ref.c (split_constant_offset_1): Likewise.
1603         (dr_analyze_innermost): Likewise.
1604         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1605         * tree-sra.c (ipa_sra_check_caller): Likewise.
1606         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1607         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1608         (instrument_bool_enum_load, instrument_object_size): Likewise.
1609         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1610         to get_inner_reference.
1611         * hsa-gen.c (gen_hsa_addr): Likewise.
1612         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1613         * tsan.c (instrument_expr): Likewise.
1614         * match.pd: Update call to ptr_difference_const.
1616 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1617             Alan Hayward  <alan.hayward@arm.com>
1618             David Sherwood  <david.sherwood@arm.com>
1620         * fold-const.c (fold_comparison): Track sizes and offsets as
1621         poly_int64s rather than HOST_WIDE_INTs when folding address
1622         comparisons.
1624 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1625             Alan Hayward  <alan.hayward@arm.com>
1626             David Sherwood  <david.sherwood@arm.com>
1628         * expr.h (get_bit_range): Return the bitstart and bitend as
1629         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1630         as a poly_int64 rather than a HOST_WIDE_INT.
1631         * expr.c (get_bit_range): Likewise.
1632         (expand_assignment): Update call accordingly.
1633         * fold-const.c (optimize_bit_field_compare): Likewise.
1635 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1636             Alan Hayward  <alan.hayward@arm.com>
1637             David Sherwood  <david.sherwood@arm.com>
1639         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1640         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1641         as an optional final parameter.
1642         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1643         (aarch64_legitimate_address_p): Likewise.
1644         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1645         instead of an rtx code.
1646         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1647         (aarch64_legitimate_address_hook_p): Likewise.
1648         (aarch64_print_ldpstp_address): Likewise.
1649         (aarch64_print_operand_address): Likewise.
1650         (aarch64_address_cost): Likewise.
1651         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1652         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1653         (aarch64_mem_pair_lanes_operand): Likewise.
1655 2017-12-20  Richard Biener  <rguenther@suse.de>
1657         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1658         update_stmt_if_modified.
1660 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1662         PR tree-optimization/83491
1663         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1664         before walking uses.  Improve coding style and comments.
1666 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1668         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1670 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1672         PR ipa/83506
1673         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1674         todo_flags_finish.
1675         (pass_ipa_free_fn_summary): Add small_p private data member,
1676         initialize to false in the ctor.
1677         (pass_ipa_free_fn_summary::clone,
1678         pass_ipa_free_fn_summary::set_pass_param,
1679         pass_ipa_free_fn_summary::gate): New methods.
1680         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1681         | TODO_dump_symtab if small_p.
1682         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1683         entry and add another instance of the pass with false parm after
1684         ipa-pure-const.
1685         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1686         ipa_free_fn_summary here.
1688 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1690         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1692 2017-12-20  Martin Sebor  <msebor@redhat.com>
1694         PR testsuite/83131
1695         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1696         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1697         (expand_builtin_strncmp): Same.
1699 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1701         PR bootstrap/83396
1702         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1703         allowed debug stmts before labels.
1704         (expand_gimple_basic_block): Likewise.
1705         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1706         * gimple-iterator.h (gsi_after_labels): Likewise.
1707         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1708         rename reused variable, and simplify using gsi_move_before.
1709         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1710         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1711         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1712         (gimple_verify_flow_info, gimple_block_label): Likewise.
1713         (make_blocks): Move debug markers after adjacent labels.
1714         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1715         allowed debug insns outside blocks.
1716         * df-scan.c (df_insn_delete): Likewise.
1717         * lra-constraints.c (update_ebb_live_info): Likewise.
1718         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1719         (vt_initialize, delete_vta_debug_insns): Likewise.
1720         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1722 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1723             Alan Hayward  <alan.hayward@arm.com>
1724             David Sherwood  <david.sherwood@arm.com>
1726         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1727         (round_up_to_byte_boundary): Likewise.
1728         * expr.h (get_bit_range): Add temporary shim.
1729         * gimple-ssa-store-merging.c (store_operand_info): Change the
1730         bitsize, bitpos, bitregion_start and bitregion_end fields from
1731         unsigned HOST_WIDE_INT to poly_uint64.
1732         (merged_store_group): Likewise load_align_base.
1733         (compatible_load_p, compatible_load_p): Update accordingly.
1734         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1735         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1736         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1737         bitregion_start and bitregion_end as poly_uint64s rather than
1738         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1739         (handled_load): Take the bitsize, bitpos,
1740         bitregion_start and bitregion_end as poly_uint64s rather than
1741         unsigned HOST_WIDE_INTs.
1742         (pass_store_merging::process_store): Update call to
1743         mem_valid_for_store_merging.
1745 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1746             Alan Hayward  <alan.hayward@arm.com>
1747             David Sherwood  <david.sherwood@arm.com>
1749         * builtins.c (get_object_alignment_2): Track polynomial offsets
1750         and sizes.  Update the alignment handling.
1752 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1753             Alan Hayward  <alan.hayward@arm.com>
1754             David Sherwood  <david.sherwood@arm.com>
1756         * tree.h (get_inner_reference): Add a version that returns the
1757         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1758         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1759         the case in which bitpos is not associated with the first byte.
1761 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1762             Alan Hayward  <alan.hayward@arm.com>
1763             David Sherwood  <david.sherwood@arm.com>
1765         * tree-affine.h (get_inner_reference_aff): Return the size as a
1766         poly_widest_int.
1767         * tree-affine.c (get_inner_reference_aff): Likewise.
1768         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1769         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1771 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1772             Alan Hayward  <alan.hayward@arm.com>
1773             David Sherwood  <david.sherwood@arm.com>
1775         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1776         HOST_WIDE_INT rather than a poly_int64.
1778 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1779             Alan Hayward  <alan.hayward@arm.com>
1780             David Sherwood  <david.sherwood@arm.com>
1782         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1783         HOST_WIDE_INT to poly_int64_pod.
1784         (perform_symbolic_merge): Update accordingly.
1785         (bswap_replace): Likewise.
1787 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1788             Alan Hayward  <alan.hayward@arm.com>
1789             David Sherwood  <david.sherwood@arm.com>
1791         * tree-affine.h (aff_tree::offset): Change from widest_int
1792         to poly_widest_int.
1793         (wide_int_ext_for_comb): Delete.
1794         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1795         constants as poly_widest_int rather than widest_int.
1796         (aff_combination_constant_multiple_p): Return the multiplier
1797         as a poly_widest_int.
1798         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1799         polynomial offsets.
1800         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1801         version static and add an overload for poly_widest_int.
1802         (aff_combination_const, aff_combination_add_cst)
1803         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1804         the constants as poly_widest_int rather than widest_int.
1805         (tree_to_aff_combination): Generalize INTEGER_CST case to
1806         poly_int_tree_p.
1807         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1808         (aff_combination_add_product, aff_combination_mult): Handle
1809         polynomial offsets.
1810         (aff_combination_constant_multiple_p): Return the multiplier
1811         as a poly_widest_int.
1812         * tree-predcom.c (determine_offset): Return the offset as a
1813         poly_widest_int.
1814         (split_data_refs_to_components, suitable_component_p): Update
1815         accordingly.
1816         (valid_initializer_p): Update call to
1817         aff_combination_constant_multiple_p.
1818         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1819         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1820         as a poly_int64 rather than a HOST_WIDE_INT.
1821         (get_address_cost): Handle polynomial offsets.
1822         (iv_elimination_compare_lt): Likewise.
1823         (rewrite_use_nonlinear_expr): Likewise.
1825 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1826             Alan Hayward  <alan.hayward@arm.com>
1827             David Sherwood  <david.sherwood@arm.com>
1829         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1830         as a poly_int64_pod rather than a HOST_WIDE_INT.
1831         (get_addr_base_and_unit_offset): Likewise.
1832         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1833         (get_addr_base_and_unit_offset): Likewise.
1834         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1835         to poly_int64 in example.
1836         * fold-const.c (fold_binary_loc): Update call to
1837         get_addr_base_and_unit_offset.
1838         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1839         (maybe_canonicalize_mem_ref_addr): Likewise.
1840         (gimple_fold_stmt_to_constant_1): Likewise.
1841         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1842         Likewise.
1843         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1844         * match.pd: Likewise.
1845         * omp-low.c (lower_omp_target): Likewise.
1846         * tree-sra.c (build_ref_for_offset): Likewise.
1847         (build_debug_ref_for_model): Likewise.
1848         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1849         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1850         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1851         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1852         (constant_pointer_difference): Likewise.
1853         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1854         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1855         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1856         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1857         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1858         (set_ssa_val_to): Likewise.
1859         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1860         (maybe_diag_stxncpy_trunc): Likewise.
1861         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1862         * tree.c (build_simple_mem_ref_loc): Likewise.
1863         (array_at_struct_end_p): Likewise.
1865 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1866             Alan Hayward  <alan.hayward@arm.com>
1867             David Sherwood  <david.sherwood@arm.com>
1869         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
1870         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1871         (get_ref_base_and_extent_hwi): Declare.
1872         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
1873         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1874         (get_ref_base_and_extent_hwi): New function.
1875         * cfgexpand.c (expand_debug_expr): Update call to
1876         get_ref_base_and_extent.
1877         * dwarf2out.c (add_var_loc_to_decl): Likewise.
1878         * gimple-fold.c (get_base_constructor): Return the offset as a
1879         poly_int64_pod rather than a HOST_WIDE_INT.
1880         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
1881         * ipa-polymorphic-call.c
1882         (ipa_polymorphic_call_context::set_by_invariant)
1883         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
1884         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
1885         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
1886         rather than get_ref_base_and_extent.
1887         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
1888         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
1889         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
1890         Likewise.
1891         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
1892         call to get_ref_base_and_extent.
1893         * tree-sra.c (create_access, get_access_for_expr): Likewise.
1894         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
1895         (stmt_kills_ref_p): Likewise.
1896         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
1897         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
1898         Likewise.
1899         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
1900         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
1901         when calling native_encode_expr.
1902         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
1903         call to get_ref_base_and_extent.
1904         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
1905         get_ref_base_and_extent.
1906         * var-tracking.c (track_expr_p): Likewise.
1908 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1909             Alan Hayward  <alan.hayward@arm.com>
1910             David Sherwood  <david.sherwood@arm.com>
1912         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
1913         HOST_WIDE_INT to poly_int64_pod.
1914         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
1915         polynomail parameter offsets.
1917 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1918             Alan Hayward  <alan.hayward@arm.com>
1919             David Sherwood  <david.sherwood@arm.com>
1921         * gengtype.c (main): Handle poly_int64_pod.
1922         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
1923         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
1924         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
1925         from HOST_WIDE_INT to poly_int64_pod.
1926         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
1927         (copy_cfa): New function.
1928         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
1929         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
1930         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
1931         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
1932         (initial_return_save): Treat offsets as poly_ints.
1933         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
1934         offset is nonconstant.
1935         (reg_save): Take the offset as a poly_int64.  Fall back to
1936         DW_CFA_expression for nonconstant offsets.
1937         (queue_reg_save): Take the offset as a poly_int64.
1938         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
1940 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1941             Alan Hayward  <alan.hayward@arm.com>
1942             David Sherwood  <david.sherwood@arm.com>
1944         * rtl.h (operand_subword, operand_subword_force): Take the offset
1945         as a poly_uint64 an unsigned int.
1946         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
1948 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1949             Alan Hayward  <alan.hayward@arm.com>
1950             David Sherwood  <david.sherwood@arm.com>
1952         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
1953         'p' format code.  Use INT_LIST rather than SUBREG as the example of
1954         a code with an XINT and an XEXP.  Remove the implication that
1955         accessing an rtx field using XINT is expected to work.
1956         * rtl.def (SUBREG): Change format from "ei" to "ep".
1957         * rtl.h (rtunion::rt_subreg): New field.
1958         (XCSUBREG): New macro.
1959         (SUBREG_BYTE): Use it.
1960         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
1961         Update constructor accordingly.
1962         (subreg_shape::operator ==): Update accordingly.
1963         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
1964         than an unsigned int.
1965         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
1966         a poly_uint64 rather than an unsigned int.
1967         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
1968         than an unsigned int.
1969         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
1970         (subreg_size_highpart_offset): Return a poly_uint64 rather than
1971         an unsigned int.  Take the sizes as poly_uint64s.
1972         (subreg_offset_from_lsb): Return a poly_uint64 rather than
1973         an unsigned int.  Take the shift as a poly_uint64 rather than
1974         an unsigned int.
1975         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
1976         as a poly_uint64 rather than an unsigned int.
1977         (simplify_subreg_regno): Likewise.
1978         (byte_lowpart_offset): Return the memory offset as a poly_int64
1979         rather than an int.
1980         (subreg_memory_offset): Likewise.  Take the subreg offset as a
1981         poly_uint64 rather than an unsigned int.
1982         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
1983         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
1984         poly_uint64 rather than an unsigned int.
1985         * rtl.c (rtx_format): Describe 'p' in comment.
1986         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
1987         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
1988         offset as a poly_uint64 rather than an unsigned int.
1989         (byte_lowpart_offset): Return the memory offset as a poly_int64
1990         rather than an int.
1991         (subreg_memory_offset): Likewise.  Take the subreg offset as a
1992         poly_uint64 rather than an unsigned int.
1993         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
1994         mode sizes as poly_uint64s rather than unsigned ints.  Return a
1995         poly_uint64 rather than an unsigned int.
1996         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
1997         (copy_insn_1): Handle 'p'.
1998         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
1999         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2000         an unsigned int.  Return the shift in the same way.
2001         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2002         unsigned int.
2003         (subreg_size_offset_from_lsb): Take the sizes and shift as
2004         poly_uint64s rather than unsigned ints.  Return the offset as
2005         a poly_uint64.
2006         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2007         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2008         an unsigned int.
2009         * rtlhash.c (add_rtx): Handle 'p'.
2010         * genemit.c (gen_exp): Likewise.
2011         * gengenrtl.c (type_from_format, gendef): Likewise.
2012         * gensupport.c (subst_pattern_match, get_alternatives_number)
2013         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2014         (subst_dup): Likewise.
2015         * gengtype.c (adjust_field_rtx_def): Likewise.
2016         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2017         (match_pattern_2): Likewise.
2018         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2019         (rtx_test::subreg_field): New function.
2020         (operator ==, safe_to_hoist_p, transition_parameter_type)
2021         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2022         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2023         * genpeep.c (match_rtx): Likewise.
2024         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2025         (rtx_writer::print_rtx_operand): Handle 'p'.
2026         (print_value): Handle SUBREG.
2027         * read-rtl.c (apply_int_iterator): Likewise.
2028         (rtx_reader::read_rtx_operand): Handle 'p'.
2029         * alias.c (rtx_equal_for_memref_p): Likewise.
2030         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2031         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2032         as poly_ints.
2033         * calls.c (expand_call): Likewise.
2034         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2035         (make_extraction, gen_lowpart_for_combine): Likewise.
2036         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2037         Likewise.
2038         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2039         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2040         (exp_equiv_p): Handle 'p'.
2041         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2042         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2043         * dse.c (find_shift_sequence): Likewise.
2044         * dwarf2out.c (rtl_for_decl_location): Likewise.
2045         * expmed.c (extract_low_bits): Likewise.
2046         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2047         (expand_expr_real_2): Likewise.
2048         * final.c (alter_subreg): Likewise.
2049         (leaf_renumber_regs_insn): Handle 'p'.
2050         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2051         Treat subreg offsets as poly_ints.
2052         * fwprop.c (forward_propagate_and_simplify): Likewise.
2053         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2054         * ira.c (get_subreg_tracking_sizes): Likewise.
2055         * ira-conflicts.c (go_through_subreg): Likewise.
2056         * ira-lives.c (process_single_reg_class_operands): Likewise.
2057         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2058         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2059         as a poly_uint64 rather than an unsigned int.
2060         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2061         subreg offsets as poly_ints.
2062         * lra-constraints.c (operands_match_p): Handle 'p'.
2063         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2064         * lra-spills.c (assign_mem_slot): Likewise.
2065         * postreload.c (move2add_valid_value_p): Likewise.
2066         * recog.c (general_operand, indirect_operand): Likewise.
2067         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2068         (copyprop_hardreg_forward_1): Likewise.
2069         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2070         (record_subregs_of_mode): Likewise.
2071         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2072         * reload.c (operands_match_p): Handle 'p'.
2073         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2074         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2075         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2076         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2077         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2078         (simplify_const_poly_int_tests<N>::run): Likewise.
2079         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2080         a poly_uint64 rather than an unsigned int.
2081         * valtrack.c (debug_lowpart_subreg): Likewise.
2082         * var-tracking.c (var_lowpart): Likewise.
2083         (loc_cmp): Handle 'p'.
2085 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2086             Alan Hayward  <alan.hayward@arm.com>
2087             David Sherwood  <david.sherwood@arm.com>
2089         * ira.c (get_subreg_tracking_sizes): New function.
2090         (init_live_subregs): Take an integer size rather than a register.
2091         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2092         to init_live_subregs.
2094 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2095             Alan Hayward  <alan.hayward@arm.com>
2096             David Sherwood  <david.sherwood@arm.com>
2098         * expr.c (store_constructor_field): Change bitsize from a
2099         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2100         HOST_WIDE_INT to a poly_int64.
2101         (store_constructor): Change size from a HOST_WIDE_INT to
2102         a poly_int64.
2103         (store_field): Likewise bitsize and bitpos.
2105 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2106             Alan Hayward  <alan.hayward@arm.com>
2107             David Sherwood  <david.sherwood@arm.com>
2109         * expmed.h (store_bit_field): Change bitregion_start and
2110         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2111         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2112         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2113         (store_fixed_bit_field, store_split_bit_field): Likewise.
2114         * expr.c (store_constructor_field, store_field): Likewise.
2115         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2116         to bitsize and bitpos.
2117         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2118         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2119         same change in the constructor arguments.
2120         (get_best_mode): Change bitregion_start and bitregion_end from
2121         unsigned HOST_WIDE_INT to poly_uint64.
2122         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2123         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2124         poly_int64.
2125         (bit_field_mode_iterator::next_mode): Update for new types
2126         of m_bitregion_start and m_bitregion_end.
2127         (get_best_mode): Change bitregion_start and bitregion_end from
2128         unsigned HOST_WIDE_INT to poly_uint64.
2130 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2131             Alan Hayward  <alan.hayward@arm.com>
2132             David Sherwood  <david.sherwood@arm.com>
2134         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2135         accepts poly_uint64 offsets.
2136         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2137         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2138         * expmed.c (lowpart_bit_field_p): Likewise.
2139         (extract_bit_field_as_subreg): New function, split out from...
2140         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2141         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2142         extractions, check that BITSIZE matches the size of the extracted
2143         value and that BITNUM is an exact multiple of that size.
2144         If all else fails, try forcing the value into memory if
2145         BITNUM is variable, and adjusting the address so that the
2146         offset is constant.  Split the part that can only handle constant
2147         bitsize and bitnum out into...
2148         (extract_integral_bit_field): ...this new function.
2149         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2150         rather than unsigned HOST_WIDE_INTs.
2152 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2153             Alan Hayward  <alan.hayward@arm.com>
2154             David Sherwood  <david.sherwood@arm.com>
2156         * expmed.h (store_bit_field): Take bitsize and bitnum as
2157         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2158         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2159         that returns the byte size.
2160         (store_bit_field_1): Take bitsize and bitnum as
2161         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2162         to simple_mem_bitfield_p.  Split the part that can only handle
2163         constant bitsize and bitnum out into...
2164         (store_integral_bit_field): ...this new function.
2165         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2166         than unsigned HOST_WIDE_INTs.
2167         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2169 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2170             Alan Hayward  <alan.hayward@arm.com>
2171             David Sherwood  <david.sherwood@arm.com>
2173         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2174         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2175         to poly_int64.
2176         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2177         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2178         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2179         offset as a poly_int64 rather than an int.
2180         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2181         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2182         * lra-constraints.c (equiv_address_substitution): Track offsets
2183         as poly_int64s.
2184         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2185         (curr_insn_transform): Handle the new form of sp_offset.
2186         * lra-eliminations.c (lra_elim_table): Change previous_offset
2187         and offset from HOST_WIDE_INT to poly_int64.
2188         (print_elim_table, update_reg_eliminate): Update accordingly.
2189         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2190         (get_elimination): Update accordingly.
2191         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2192         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2193         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2194         poly_int64 offsets generally.
2195         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2196         (mark_not_eliminable, init_elimination): Update accordingly.
2197         (remove_reg_equal_offset_note): Return a bool and pass the new
2198         offset back by pointer as a poly_int64.
2199         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2200         rather than a HOST_WIDE_INT.
2201         (do_remat): Track offsets poly_int64s.
2202         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2204 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2205             Alan Hayward  <alan.hayward@arm.com>
2206             David Sherwood  <david.sherwood@arm.com>
2208         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2209         offset from HOST_WIDE_INT to poly_int64.
2210         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2211         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2212         (widen_memory_access): Take the sizes and offsets as poly_int64s
2213         rather than HOST_WIDE_INTs.
2214         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2215         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2216         and ints.
2217         (adjust_offset_for_component_ref): Change the offset from a
2218         HOST_WIDE_INT to a poly_int64.
2219         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2220         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2221         * dce.c (find_call_stack_args): Likewise.
2222         * dse.c (record_store): Likewise.
2223         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2224         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2225         * read-rtl-function.c (test_loading_mem): Likewise.
2226         * rtlanal.c (may_trap_p_1): Likewise.
2227         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2228         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2229         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2230         (mem_attrs::mem_attrs): New function.
2231         (set_mem_attributes_minus_bitpos): Change bitpos from a
2232         HOST_WIDE_INT to poly_int64.
2233         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2234         (clear_mem_offset, clear_mem_size, change_address)
2235         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2236         initialize mem_attrs.
2237         (set_mem_offset, set_mem_size, adjust_address_1)
2238         (adjust_automodify_address_1, offset_address, widen_memory_access):
2239         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2241 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2242             Alan Hayward  <alan.hayward@arm.com>
2243             David Sherwood  <david.sherwood@arm.com>
2245         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2246         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2247         rather than 0 to represent an unknown size.  Assert that the size
2248         is known when the mode isn't BLKmode.
2249         (may_trap_p_1): Use -1 for unknown sizes.
2250         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2252 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2253             Alan Hayward  <alan.hayward@arm.com>
2254             David Sherwood  <david.sherwood@arm.com>
2256         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2257         to poly_int64.  Update commentary for positions_needed.large.
2258         (read_info_type): Change offset and width from HOST_WIDE_INT
2259         to poly_int64.
2260         (set_usage_bits): Likewise.
2261         (canon_address): Return the offset as a poly_int64 rather than
2262         a HOST_WIDE_INT.  Use strip_offset_and_add.
2263         (set_all_positions_unneeded, any_positions_needed_p): Use
2264         positions_needed.large to track stores with non-constant widths.
2265         (all_positions_needed_p): Likewise.  Take the offset and width
2266         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2267         (record_store): Cope with non-constant offsets and widths.
2268         Nullify the rhs of an earlier store if we can't tell which bytes
2269         of it are needed.
2270         (find_shift_sequence): Take the access_size and shift as poly_int64s
2271         rather than ints.
2272         (get_stored_val): Take the read_offset and read_width as poly_int64s
2273         rather than HOST_WIDE_INTs.
2274         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2275         non-constant offsets and widths.
2277 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2278             Alan Hayward  <alan.hayward@arm.com>
2279             David Sherwood  <david.sherwood@arm.com>
2281         * inchash.h (inchash::hash::add_poly_int): New function.
2282         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2283         Use poly_int64 rather than HOST_WIDE_INT.
2284         (ao_ref::max_size_known_p): New function.
2285         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2286         rather than HOST_WIDE_INT.
2287         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2288         to temporaries until its interface is adjusted to match.
2289         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2290         (aliasing_component_refs_p, decl_refs_may_alias_p)
2291         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2292         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2293         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2294         ao_ref fields.
2295         * alias.c (ao_ref_from_mem): Likewise.
2296         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2297         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2298         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2299         (maybe_trim_complex_store, maybe_trim_constructor_store)
2300         (live_bytes_read, dse_classify_store): Likewise.
2301         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2302         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2303         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2304         (vn_reference_lookup_3): Likewise.
2305         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2307 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2308             Alan Hayward  <alan.hayward@arm.com>
2309             David Sherwood  <david.sherwood@arm.com>
2311         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2312         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2313         instead of ranges_overlap_p.
2315 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2316             Alan Hayward  <alan.hayward@arm.com>
2317             David Sherwood  <david.sherwood@arm.com>
2319         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2320         sizes as poly_int64s rather than HOST_WIDE_INTs.
2322 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2323             Alan Hayward  <alan.hayward@arm.com>
2324             David Sherwood  <david.sherwood@arm.com>
2326         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2327         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2328         * gimple-fold.c (fold_ctor_reference): Likewise.
2330 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2331             Alan Hayward  <alan.hayward@arm.com>
2332             David Sherwood  <david.sherwood@arm.com>
2334         * target.def (dwarf_poly_indeterminate_value): New hook.
2335         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2336         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2337         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2338         * doc/tm.texi: Regenerate.
2339         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2340         offset as a poly_int64.
2341         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2342         offset as a poly_int64.
2343         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2344         Take the offset as a poly_int64.
2345         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2346         (frame_pointer_fb_offset): Change to a poly_int64.
2347         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2348         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2349         (based_loc_descr): Take the offset as a poly_int64.
2350         Use strip_offset_and_add to handle (plus X (const)).
2351         Use new_reg_loc_descr instead of an open-coded version of the
2352         previous implementation.
2353         (mem_loc_descriptor): Handle CONST_POLY_INT.
2354         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2355         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2357 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2358             Alan Hayward  <alan.hayward@arm.com>
2359             David Sherwood  <david.sherwood@arm.com>
2361         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2362         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2363         * inchash.h (inchash::hash::add_poly_hwi): New function.
2364         * gengtype.c (main): Register poly_int64.
2365         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2366         offset as a poly_int.
2367         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2368         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2369         offset as a poly_int64.
2370         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2371         * print-rtl.c (print_poly_int): New function.
2372         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2373         a poly_int.
2374         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2375         functions.
2376         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2377         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2378         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2379         Enforce track_offset_p for parts of a PARALLEL.
2380         (vt_add_function_parameter): Use const_offset for the final
2381         offset to track.  Use get_tracked_reg_offset for the parts
2382         of a PARALLEL.
2384 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2385             Alan Hayward  <alan.hayward@arm.com>
2386             David Sherwood  <david.sherwood@arm.com>
2388         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2389         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2390         * doc/tm.texi: Regenerate.
2391         * hooks.h (hook_bool_uint_uint_true): Delete.
2392         (hook_bool_puint64_puint64_true): Declare.
2393         * hooks.c (hook_bool_uint_uint_true): Delete.
2394         (hook_bool_puint64_puint64_true): New function.
2395         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2396         instead of unsigned ints.
2397         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2398         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2400 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2401             Alan Hayward  <alan.hayward@arm.com>
2402             David Sherwood  <david.sherwood@arm.com>
2404         * optabs.h (expand_operand): Add an int_value field.
2405         (create_expand_operand): Add an int_value parameter and use it
2406         to initialize the new expand_operand field.
2407         (create_integer_operand): Replace with a declaration of a function
2408         that accepts poly_int64s.  Move the implementation to...
2409         * optabs.c (create_integer_operand): ...here.
2410         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2411         the mode preserves the value of int_value, instead of calling
2412         const_int_operand on the rtx.  Use gen_int_mode to generate
2413         the new rtx.
2415 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2416             Alan Hayward  <alan.hayward@arm.com>
2417             David Sherwood  <david.sherwood@arm.com>
2419         * dumpfile.h (dump_dec): Declare.
2420         * dumpfile.c (dump_dec): New function.
2421         * pretty-print.h (pp_wide_integer): Turn into a function and
2422         declare a poly_int version.
2423         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2425 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2426             Alan Hayward  <alan.hayward@arm.com>
2427             David Sherwood  <david.sherwood@arm.com>
2429         * doc/generic.texi (POLY_INT_CST): Document.
2430         * tree.def (POLY_INT_CST): New tree code.
2431         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2432         * tree-core.h (tree_poly_int_cst): New struct.
2433         (tree_node): Add a poly_int_cst field.
2434         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2435         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2436         instead of a wide_int_ref.
2437         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2438         of a HOST_WIDE_INT.
2439         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2440         instead of an unsigned HOST_WIDE_INT.
2441         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2442         (ptrdiff_tree_p): Declare.
2443         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2444         extern inline implementations if the target doesn't use POLY_INT_CST.
2445         (poly_int_tree_p): New function.
2446         (wi::unextended_tree): New class.
2447         (wi::int_traits <unextended_tree>): New override.
2448         (wi::extended_tree): Add a default constructor.
2449         (wi::extended_tree::get_tree): New function.
2450         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2451         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2452         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2453         (wi::tree_to_poly_wide_ref): New typedefs.
2454         (wi::ints_for): Provide overloads for extended_tree and
2455         unextended_tree.
2456         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2457         (wi::to_wide): New functions.
2458         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2459         * tree.c (poly_int_cst_hasher): New struct.
2460         (poly_int_cst_hash_table): New variable.
2461         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2462         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2463         POLY_INT_CST.
2464         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2465         (init_ttree): Initialize poly_int_cst_hash_table.
2466         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2467         a poly_int64 instead of a HOST_WIDE_INT.
2468         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2469         instead of an unsigned HOST_WIDE_INT.
2470         (wide_int_to_tree): Rename to...
2471         (wide_int_to_tree_1): ...this.
2472         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2473         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2474         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2475         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2476         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2477         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2478         TS_POLY_INT_CST.
2479         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2480         (streamer_read_tree_body): Likewise.
2481         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2482         (streamer_write_tree_body): Likewise.
2483         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2484         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2485         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2486         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2487         * gimple-expr.h (is_gimple_constant): Likewise.
2488         * gimplify.c (maybe_with_size_expr): Likewise.
2489         * print-tree.c (print_node): Likewise.
2490         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2491         * tree-pretty-print.c (dump_generic_node): Likewise.
2492         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2493         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2494         * tree-vrp.c (compare_values_warnv): Likewise.
2495         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2496         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2497         (force_expr_to_var_cost): Likewise.
2498         * tree-ssa-loop.c (for_each_index): Likewise.
2499         * fold-const.h (build_invariant_address, size_int_kind): Take a
2500         poly_int64 instead of a HOST_WIDE_INT.
2501         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2502         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2503         POLY_INT_CST.
2504         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2505         (int_const_binop_2): New function, split out from...
2506         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2507         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2508         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2509         * gimple-ssa-strength-reduction.c (slsr_process_add)
2510         (slsr_process_mul): Check for INTEGER_CSTs before using them
2511         as candidates.
2512         * stor-layout.c (bits_from_bytes): New function.
2513         (bit_from_pos): Use it.
2514         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2515         by BITS_PER_UNIT to get the TYPE_SIZE.
2516         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2517         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2519 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2520             Alan Hayward  <alan.hayward@arm.com>
2521             David Sherwood  <david.sherwood@arm.com>
2523         * doc/rtl.texi (const_poly_int): Document.  Also document the
2524         rtl sharing behavior.
2525         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2526         * rtl.h (const_poly_int_def): New struct.
2527         (rtx_def::u): Add a cpi field.
2528         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2529         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2530         (wi::rtx_to_poly_wide_ref): New typedef
2531         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2532         (poly_int_rtx_p): New functions.
2533         (trunc_int_for_mode): Declare a poly_int64 version.
2534         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2535         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2536         a wide_int_ref.
2537         (strip_offset): Declare.
2538         (strip_offset_and_add): New function.
2539         * rtl.def (CONST_POLY_INT): New rtx code.
2540         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2541         (shared_const_p): Use poly_int_rtx_p.
2542         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2543         HOST_WIDE_INT.
2544         (gen_int_shift_amount): Likewise.
2545         * emit-rtl.c (const_poly_int_hasher): New class.
2546         (const_poly_int_htab): New variable.
2547         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2548         (const_poly_int_hasher::hash): New function.
2549         (const_poly_int_hasher::equal): Likewise.
2550         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2551         (immed_wide_int_const): Rename to...
2552         (immed_wide_int_const_1): ...this and make static.
2553         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2554         instead of a wide_int_ref.
2555         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2556         (gen_lowpart_common): Handle CONST_POLY_INT.
2557         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2558         * cselib.c (cselib_hash_rtx): Likewise.
2559         * dwarf2out.c (const_ok_for_output_1): Likewise.
2560         * expr.c (convert_modes): Likewise.
2561         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2562         * rtlhash.c (add_rtx): Likewise.
2563         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2564         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2565         Handle existing CONST_POLY_INT rtxes.
2566         * expmed.h (expand_shift): Take a poly_int64 instead of a
2567         HOST_WIDE_INT.
2568         * expmed.c (expand_shift): Likewise.
2569         * rtlanal.c (strip_offset): New function.
2570         (commutative_operand_precedence): Give CONST_POLY_INT the same
2571         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2572         and CONST_INT.
2573         * rtl-tests.c (const_poly_int_tests): New struct.
2574         (rtl_tests_c_tests): Use it.
2575         * simplify-rtx.c (simplify_const_unary_operation): Handle
2576         CONST_POLY_INT.
2577         (simplify_const_binary_operation): Likewise.
2578         (simplify_binary_operation_1): Fold additions of symbolic constants
2579         and CONST_POLY_INTs.
2580         (simplify_subreg): Handle extensions and truncations of
2581         CONST_POLY_INTs.
2582         (simplify_const_poly_int_tests): New struct.
2583         (simplify_rtx_c_tests): Use it.
2584         * wide-int.h (storage_ref): Add default constructor.
2585         (wide_int_ref_storage): Likewise.
2586         (trailing_wide_ints): Use GTY((user)).
2587         (trailing_wide_ints::operator[]): Add a const version.
2588         (trailing_wide_ints::get_precision): New function.
2589         (trailing_wide_ints::extra_size): Likewise.
2591 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2592             Alan Hayward  <alan.hayward@arm.com>
2593             David Sherwood  <david.sherwood@arm.com>
2595         * emit-rtl.h (gen_int_shift_amount): Declare.
2596         * emit-rtl.c (gen_int_shift_amount): New function.
2597         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2598         instead of GEN_INT.
2599         * calls.c (shift_return_value): Likewise.
2600         * cse.c (fold_rtx): Likewise.
2601         * dse.c (find_shift_sequence): Likewise.
2602         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2603         (expand_shift, expand_smod_pow2): Likewise.
2604         * lower-subreg.c (shift_cost): Likewise.
2605         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2606         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2607         (expand_vec_perm_var): Likewise.
2608         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2609         (simplify_binary_operation_1): Likewise.
2610         * combine.c (try_combine, find_split_point, force_int_to_mode)
2611         (simplify_shift_const_1, simplify_shift_const): Likewise.
2612         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2614 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2616         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2618 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2620         * doc/invoke.texi (ARM Options): Document accepted extension options
2621         for -march=armv8.3-a.
2623 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2625         PR target/83105
2626         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2627         or --with-float=softfp, set the default CPU to arm10e.
2629 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2631         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2632         * config/visium/predicates.md (const_shift_operand): Likewise.
2633         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2634         (visium_legitimize_reload_address): Likewise.
2636 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2638         PR target/82975
2639         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2641 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2643         PR c++/83490
2644         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2646 2017-12-20  Julia Koval  <julia.koval@intel.com>
2648         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2649         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2650         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2651         * config.gcc: Include vpclmulqdqintrin.h.
2652         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2653         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2654         -mvpclmulqdq.
2655         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2656         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2657         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2658         * config/i386/i386.c (isa2_opts): Add -mcx16.
2659         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2660         (ix86_option_override_internal): Move mcx16 to flags2.
2661         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2662         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2663         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2664         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2665         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2666         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2667         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2668         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2669         * doc/invoke.texi: Add -mvpclmulqdq.
2671 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2673         PR middle-end/83423
2674         * config/i386/i386.c (ix86_static_chain): Move
2675         DECL_STATIC_CHAIN test ...
2676         * calls.c (rtx_for_static_chain): ... here.  New function.
2677         * calls.h (rtx_for_static_chain): Declare.
2678         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2679         instead of targetm.calls.static_chain.
2680         * df-scan.c (df_get_entry_block_def_set): Same.
2682 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2684         PR tree-optimization/83493
2685         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2686         and lb.
2688 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2690         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2691         inform with hardcoded english plural handling.
2693 2017-12-18  Jeff Law  <law@redhat.com>
2695         PR tree-optimization/83477
2696         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2697         a non-virtual PHI, always push a new range.
2699 2017-12-19  Martin Sebor  <msebor@redhat.com>
2701         PR middle-end/77608
2702         * builtins.c (compute_objsize): Handle non-constant offsets.
2704 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2706         PR tree-optimization/83444
2707         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2708         character load case, if get_stridx on MEM_REF's operand doesn't
2709         look usable, retry with get_addr_stridx.
2711 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2713         PR debug/83422
2714         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2715         (variable_tracking_main_1): Keep markers even when VTA fails.
2717         PR bootstrap/83396
2718         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2719         even if there are markers before it.
2720         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2722 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2724         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2725         typical order conditions.
2726         * sel-sched.c (extract_new_fences_from): Likewise.
2727         * config/visium/constraints.md (J, K, L): Likewise.
2728         * config/visium/predicates.md (const_shift_operand): Likewise.
2729         * config/visium/visium.c (visium_legitimize_address,
2730         visium_legitimize_reload_address): Likewise.
2731         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2732         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2733         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2734         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2735         SET_DUMP_DETAIL): Likewise.
2736         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2737         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2738         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2739         avr_set_core_architecture, avr_set_current_function,
2740         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2741         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2742         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2743         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2744         avr_map_decompose, avr_fold_builtin): Likewise.
2745         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2746         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2747         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2748         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2749         * config/m32c/m32c.c (m32c_conditional_register_usage,
2750         m32c_address_cost): Likewise.
2751         * config/m32c/predicates.md (shiftcount_operand,
2752         longshiftcount_operand): Likewise.
2753         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2754         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2755         can_use_cdx_ldstw): Likewise.
2756         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2757         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2758         Likewise.
2759         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2760         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2761         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2762         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2763         * config/vax/vax.c (adjacent_operands_p): Likewise.
2764         * config/ft32/constraints.md (L, b, KA): Likewise.
2765         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2766         Likewise.
2767         * cfgexpand.c (expand_stack_alignment): Likewise.
2768         * gcse.c (insert_expr_in_table): Likewise.
2769         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2770         * cgraphunit.c (cgraph_node::expand): Likewise.
2771         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2772         * emit-rtl.c (add_insn): Likewise.
2773         * input.c (dump_location_info): Likewise.
2774         * passes.c (NEXT_PASS): Likewise.
2775         * read-rtl-function.c (parse_note_insn_name,
2776         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2777         Likewise.
2778         * sched-rgn.c (sched_rgn_init): Likewise.
2779         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2780         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2781         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2782         Likewise.
2783         * explow.c (eliminate_constant_term): Likewise.
2784         * final.c (leaf_renumber_regs_insn): Likewise.
2785         * cfgrtl.c (print_rtl_with_bb): Likewise.
2786         * genhooks.c (emit_init_macros): Likewise.
2787         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2788         * tree-data-ref.c (conflict_fn): Likewise.
2789         * selftest.c (assert_streq): Likewise.
2790         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2791         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2792         fold_binary_loc, multiple_of_p): Likewise.
2793         * reload.c (push_reload, find_equiv_reg): Likewise.
2794         * et-forest.c (et_nca, et_below): Likewise.
2795         * dbxout.c (dbxout_symbol_location): Likewise.
2796         * reorg.c (relax_delay_slots): Likewise.
2797         * dojump.c (do_compare_rtx_and_jump): Likewise.
2798         * gengtype-parse.c (type): Likewise.
2799         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2800         simplify_const_relational_operation): Likewise.
2801         * reload1.c (do_output_reload): Likewise.
2802         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2803         * gengtype.c (type_for_name): Likewise.
2804         * gimple-ssa-sprintf.c (format_directive): Likewise.
2806 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2808         PR target/82975
2809         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2810         accessing it.  Adjust comment.
2812 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2814         PR middle-end/81914
2815         * predict.c (zero_one_minusone): New function.
2816         (apply_return_prediction): Avoid return prediction for functions
2817         returning only -1, 0 and 1 values, unless they only return -1 and 0
2818         or 0 and 1.
2820 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2822         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2823         fall-through warning.
2824         (arc_compute_frame_size): Remove unused variables.
2825         (arc_print_operand): Fix fprintif format.
2826         (arc_can_follow_jump): Clean fall-through warning.
2828 2017-12-19  Marek Polacek  <polacek@redhat.com>
2830         PR c++/83489
2831         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2832         on an error node.
2834 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2836         * config/arc/arc.c (overriderregs): New variable.
2837         (arc_override_options): Track fixed/call saved/call options.
2838         (arc_conditional_register_usage): Check against overrideregs
2839         variable whenever we change register properties.
2841 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2843         * opts.c (finish_options): Don't prefix dump_base_name if it
2844         already contains directories.
2846 2017-12-19  Martin Liska  <mliska@suse.cz>
2848         PR rtl-optimization/82675
2849         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2850         more element in sbitmap.
2852 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
2854         * gimplify.c (gimplify_expr): Use error_operand_p.
2856 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2858         PR target/83387
2859         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2860         multilib.
2862 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
2864         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
2865         the jump is to a label.
2867 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2869         PR tree-optimization/83444
2870         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
2871         character loads.
2873         PR ipa/82801
2874         PR ipa/83346
2875         * ipa-inline.c (flatten_remove_node_hook): New function.
2876         (ipa_inline): Keep only nodes with flatten attribute at the end of
2877         the array in the order from ipa_reverse_postorder, only walk that
2878         portion of array for flattening, if there is more than one such
2879         node, temporarily register a removal hook and ignore removed nodes.
2881         PR tree-optimization/80631
2882         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
2883         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
2884         IFN_REDUC_MAX or IFN_REDUC_MIN.
2886 2017-12-18  Jeff Law  <law@redhat.com>
2888         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
2889         record symbolic equivalences from backedges in the CFG.
2891         Revert
2892         2017-11-19  Jeff Law  <law@redhat.com>
2894         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
2895         of degenerates resulting from ignoring an edge.
2897 2017-12-18  Martin Sebor  <msebor@redhat.com>
2899         PR middle-end/83373
2900         PR tree-optimization/78450
2901         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
2902         (handle_builtin_strlen): Call it.
2904 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
2906         PR rtl-optimization/83424
2907         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
2909 2017-12-18  Marek Polacek  <polacek@redhat.com>
2911         PR middle-end/83463
2912         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2913         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
2914         values.
2916 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2918         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
2919         (umaddsidi4, umaddsidi_split): Likewise.
2921 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2923         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
2924         constants.
2926 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
2928         PR target/83420
2929         * config/s390/s390.c (s390_option_override): Avoid strncpy.
2931 2017-12-18  Richard Biener  <rguenther@suse.de>
2933         PR tree-optimization/81877
2934         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
2935         (outermost_indep_loop): Adjust.
2936         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
2937         (can_sm_ref_p): Adjust.
2939 2017-12-18  Richard Biener  <rguenther@suse.de>
2941         PR middle-end/77291
2942         * tree.c (array_at_struct_end_p): Return true if the underlying
2943         object has space for at least one element in excess of what
2944         the array domain specifies.
2946 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2948         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
2949         example to avoid overfull hbox.
2950         * doc/invoke.texi (Option Summary): Add missing @gol.
2951         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
2952         overfull hbox.
2954 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2955             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2957         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
2958         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
2959         (AArch64 Options): Document -mverbose-cost-dump.
2960         (ARM Options): Likewise, plus -mflip-thumb.
2962 2017-12-17  Martin Sebor  <msebor@redhat.com>
2964         PR bootstrap/83446
2965         * gimple-ssa-warn-restrict.c
2966         (builtin_memref::offset_out_of_bounds): Correct the handling of
2967         anti-ranges.
2969 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2971         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
2973 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
2975         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
2977 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
2979         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
2980         latencies.
2982 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
2984         * doc/invoke.texi: Fix some typos.
2986 2017-12-16  Martin Sebor  <msebor@redhat.com>
2988         PR tree-optimization/78918
2989         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
2990         * builtins.c (check_sizes): Rename...
2991         (check_access): ...to this.  Rename function arguments for clarity.
2992         (check_memop_sizes): Adjust names.
2993         (expand_builtin_memchr, expand_builtin_memcpy): Same.
2994         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
2995         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
2996         (check_strncat_sizes, expand_builtin_strncat): Same.
2997         (expand_builtin_strncpy, expand_builtin_memset): Same.
2998         (expand_builtin_bzero, expand_builtin_memcmp): Same.
2999         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3000         (maybe_emit_sprintf_chk_warning): Same.
3001         (expand_builtin_strcpy): Adjust.
3002         (expand_builtin_stpcpy): Same.
3003         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3004         in pointer-checking forms of memcpy, memmove, and mempcpy.
3005         (gcall_to_tree_minimal, max_object_size): Define new functions.
3006         * builtins.h (max_object_size): Declare.
3007         * calls.c (alloc_max_size): Call max_object_size instead of
3008         hardcoding ssizetype limit.
3009         (get_size_range): Handle new argument.
3010         * calls.h (get_size_range): Add a new argument.
3011         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3012         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3013         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3014         operations.
3015         (gimple_fold_builtin_memory_chk): Same.
3016         (gimple_fold_builtin_stxcpy_chk): New function.
3017         * gimple-ssa-warn-restrict.c: New source.
3018         * gimple-ssa-warn-restrict.h: New header.
3019         * gimple.c (gimple_build_call_from_tree): Propagate location.
3020         * passes.def (pass_warn_restrict): Add new pass.
3021         * tree-pass.h (make_pass_warn_restrict): Declare.
3022         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3023         operations.
3024         (handle_builtin_strcat): Same.
3025         (strlen_optimize_stmt): Rename...
3026         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3027         stpncpy, strncpy, and their checking forms.
3029 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3031         PR rtl-optimization/82849
3032         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3033         and get_max_loop_iterations_int.
3035 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3036             Alan Hayward  <alan.hayward@arm.com>
3037             David Sherwood  <david.sherwood@arm.com>
3039         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3040         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3041         size as a poly_uint64.
3042         (mode_for_vector, mode_for_int_vector): Take the number of vector
3043         elements as a poly_uint64.
3044         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3045         size as a poly_uint64.
3046         (mode_for_vector, mode_for_int_vector): Take the number of vector
3047         elements as a poly_uint64.
3049 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3050             Alan Hayward  <alan.hayward@arm.com>
3051             David Sherwood  <david.sherwood@arm.com>
3053         * machmode.h (MACRO_MODE): New macro.
3054         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3055         * targhooks.c (default_libcall_value, default_secondary_reload)
3056         (default_memory_move_cost, default_register_move_cost)
3057         (default_class_max_nregs): Likewise.
3059 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3060             Alan Hayward  <alan.hayward@arm.com>
3061             David Sherwood  <david.sherwood@arm.com>
3063         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3064         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3065         target C file.
3066         * genautomata.c (main): Likewise.
3067         * genconditions.c (write_header): Likewise.
3068         * genemit.c (main): Likewise.
3069         * genextract.c (print_header): Likewise.
3070         * genopinit.c (main): Likewise.
3071         * genoutput.c (output_prologue): Likewise.
3072         * genpeep.c (main): Likewise.
3073         * genpreds.c (write_insn_preds_c): Likewise.
3074         * genrecog.c (writer_header): Likewise.
3075         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3076         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3077         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3078         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3079         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3080         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3081         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3082         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3083         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3084         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3085         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3086         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3087         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3088         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3089         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3090         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3091         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3092         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3093         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3094         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3095         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3096         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3097         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3098         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3099         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3100         * config/darwin.c (IN_TARGET_CODE): Likewise.
3101         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3102         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3103         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3104         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3105         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3106         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3107         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3108         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3109         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3110         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3111         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3112         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3113         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3114         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3115         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3116         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3117         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3118         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3119         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3120         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3121         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3122         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3123         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3124         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3125         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3126         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3127         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3128         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3129         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3130         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3131         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3132         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3133         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3134         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3135         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3136         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3137         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3138         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3139         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3140         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3141         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3142         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3143         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3144         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3145         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3146         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3147         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3148         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3149         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3150         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3151         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3152         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3153         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3154         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3155         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3156         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3157         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3158         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3159         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3160         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3161         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3162         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3163         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3164         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3165         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3166         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3167         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3168         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3169         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3170         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3171         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3172         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3173         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3174         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3175         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3176         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3177         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3178         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3179         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3180         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3181         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3182         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3183         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3184         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3185         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3186         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3187         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3188         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3189         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3190         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3191         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3192         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3193         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3194         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3195         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3196         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3197         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3198         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3199         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3200         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3201         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3202         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3203         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3204         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3205         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3206         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3208 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3209             Alan Hayward  <alan.hayward@arm.com>
3210             David Sherwood  <david.sherwood@arm.com>
3212         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3213         checks for MEM_REF.
3215 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3216             Alan Hayward  <alan.hayward@arm.com>
3217             David Sherwood  <david.sherwood@arm.com>
3219         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3220         * doc/md.texi (vec_series@var{m}): Document.
3221         * tree.def (VEC_SERIES_EXPR): New tree code.
3222         * tree.h (build_vec_series): Declare.
3223         * tree.c (build_vec_series): New function.
3224         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3225         * tree-pretty-print.c (dump_generic_node): Likewise.
3226         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3227         * tree-inline.c (estimate_operator_cost): Likewise.
3228         * expr.c (expand_expr_real_2): Likewise.
3229         * optabs-tree.c (optab_for_tree_code): Likewise.
3230         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3231         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3232         * expmed.c (make_tree): Handle VEC_SERIES.
3233         * optabs.def (vec_series_optab): New optab.
3234         * optabs.h (expand_vec_series_expr): Declare.
3235         * optabs.c (expand_vec_series_expr): New function.
3236         * tree-vect-generic.c (expand_vector_operations_1): Check that
3237         the operands also have vector type.
3239 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3240             Alan Hayward  <alan.hawyard@arm.com>
3241             David Sherwood  <david.sherwood@arm.com>
3243         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3244         (VEC_COND_EXPR): Add missing @tindex.
3245         * doc/md.texi (vec_duplicate@var{m}): Document.
3246         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3247         * tree.c (build_vector_from_val): Add stubbed-out handling of
3248         variable-length vectors, using VEC_DUPLICATE_EXPR.
3249         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3250         * cfgexpand.c (expand_debug_expr): Likewise.
3251         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3252         * tree-inline.c (estimate_operator_cost): Likewise.
3253         * tree-pretty-print.c (dump_generic_node): Likewise.
3254         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3255         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3256         (test_vec_duplicate_folding): New function.
3257         (fold_const_c_tests): Call it.
3258         * optabs.def (vec_duplicate_optab): New optab.
3259         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3260         * optabs.h (expand_vector_broadcast): Declare.
3261         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3262         vec_duplicate_optab.
3263         * expr.c (store_constructor): Try using vec_duplicate_optab for
3264         uniform vectors.
3265         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3267 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3269         PR target/83358
3270         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3271         div/mod latencies a bit.
3273 2017-12-15  Jeff Law  <law@redhat.com>
3275         PR tree-optimization/36550
3276         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3277         (mark_threaded_blocks): Rewrite code to avoid block copying when
3278         optimizing for size.  Don't pessimize blocks which will be
3279         copied, but all the statements will be dead.
3281 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3283         PR tree-optimization/81165
3284         * tree-ssa-threadupdate.c (uses_in_bb): New.
3285         (estimate_threading_killed_stmts): New.
3286         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3287         * tree-ssa-threadedge.c
3288         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3289         when its hit.
3291 2017-12-15  Jeff Law  <law@redhat.com>
3293         PR tree-optimization/83410
3294         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3295         threads when parallelizing loops.
3297 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3299         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3300         handler fields.
3301         * config/alpha/alpha.c (vms_attribute_table): Swap
3302         affects_type_identity and handler fields, adjust comments.
3303         * config/mips/mips.c (mips_attribute_table): Likewise.
3304         * config/visium/visium.c (visium_attribute_table): Likewise.
3305         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3306         * config/microblaze/microblaze.c (microblaze_attribute_table):
3307         Likewise.
3308         * config/spu/spu.c (spu_attribute_table): Likewise.
3309         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3310         * config/arc/arc.c (arc_attribute_table): Likewise.
3311         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3312         * config/v850/v850.c (v850_handle_interrupt_attribute,
3313         v850_handle_data_area_attribute): Formatting fixes.
3314         (v850_attribute_table): Swap affects_type_identity and handler
3315         fields, adjust comments.
3316         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3317         * config/arm/arm.c (arm_attribute_table): Likewise.
3318         * config/avr/avr.c (avr_attribute_table): Likewise.
3319         * config/s390/s390.c (s390_attribute_table): Likewise.
3320         * config/sh/sh.c (sh_attribute_table): Likewise.
3321         * config/i386/i386.c (ix86_handle_cconv_attribute,
3322         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3323         (ix86_attribute_table): Swap affects_type_identity and handler
3324         fields, adjust comments.
3325         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3326         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3327         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3328         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3329         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3330         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3331         * config/rx/rx.c (rx_attribute_table): Likewise.
3332         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3333         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3334         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3335         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3336         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3337         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3338         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3339         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3340         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3341         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3342         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3343         exclude.
3344         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3345         handler fields.
3346         (register_scoped_attributes, decl_attributes): Formatting fixes.
3348         PR tree-optimization/83269
3349         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3350         subtraction in arg0's type if type is signed and arg0 is unsigned.
3351         Formatting fix.
3353         PR sanitizer/81281
3354         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3355         convert? on @0 instead of convert.  Check type of @1, not @0.
3356         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3357         convert? on @0 instead of convert.  Check type of @1, not @0.
3358         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3359         only optimize if either both @1 and @2 types are narrower
3360         precision, or both are wider or equal precision, and in the former
3361         case only if both have undefined overflow.
3363 2017-12-15  Richard Biener  <rguenther@suse.de>
3365         PR lto/83388
3366         * internal-fn.def (IFN_NOP): Add.
3367         * internal-fn.c (expand_NOP): Do nothing.
3368         * lto-streamer-in.c (input_function): Instead of removing
3369         sanitizer calls replace them with IFN_NOP calls.
3371 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3372             Alan Hayward  <alan.hayward@arm.com>
3373             David Sherwood  <david.sherwood@arm.com>
3375         * dse.c (store_info, read_info_type): Replace begin and end with
3376         offset and width.
3377         (print_range): New function.
3378         (set_all_positions_unneeded, any_positions_needed_p)
3379         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3380         accordingly.
3381         (record_store): Likewise.  Optimize the case in which all positions
3382         are unneeded.
3383         (get_stored_val): Replace read_begin and read_end with read_offset
3384         and read_width.
3385         (replace_read): Update call accordingly.
3387 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3389         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3390         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3391         (loop_cand::loop_cand): Initialize above members.
3392         (loop_cand::supported_operations): Delete.
3393         (loop_cand::can_interchange_p): Inline above function.
3394         (loop_cand::classify_simple_reduction): Record number of constant
3395         initialized simple reductions.
3396         (should_interchange_loops): New parameters.  Check stmt cost of loops
3397         to be interchange.
3398         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3399         Update call to should_interchange_loops.
3400         (should_interchange_loop_nest): Update call to
3401         should_interchange_loops.
3403 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3405         PR target/66488
3406         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3407         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3408         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3410 2017-12-15  Julia Koval  <julia.koval@intel.com>
3412         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3413         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3414         New.
3415         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3416         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3417         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3419 2017-12-15  Julia Koval  <julia.koval@intel.com>
3421         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3422         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3423         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3424         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3425         _mm_aesenc_epi128): New intrinsics.
3427 2017-12-15  Julia Koval  <julia.koval@intel.com>
3429         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3430         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3431         New.
3432         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3433         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3434         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3436 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3438         * gimple-ssa-strength-reduction.c (analyze_increments):
3439         Distinguish replacement costs for constant strides from those for
3440         unknown strides.
3442 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3444         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3446 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3448         * doc/invoke.texi: Document -Wcast-function-type.
3449         * recog.h (stored_funcptr): Change signature.
3450         * tree-dump.c (dump_node): Avoid warning.
3451         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3453 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3455         PR middle_end/79538
3456         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3457         array.
3459 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3461         PR tree-optimization/83312
3462         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3463         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3464         "val" param, and to cope with arbitrary basic blocks.
3465         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3466         handle NULL_TREE for "val", dropping "bb" param.
3467         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3468         drop "bb" param.  Handle NULL_TREE for "val".
3469         (find_case_label_for_value): Make "switch_stmt" param const.
3470         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3471         of dom_walker.
3472         (vrp_prop::check_all_array_refs): Reimplement as...
3473         (check_array_bounds_dom_walker::before_dom_children): ...this new
3474         vfunc.  Replace linear search through BB block list, excluding
3475         those with non-executable in-edges via dominator walk.
3477 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3479         * config/arm/arm.opt (mverbose-cost-dump): New option.
3480         * config/arm/arm.c (arm_rtx_costs): Use it.
3482 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3484         PR bootstrap/83396
3485         * reload1.c (emit_input_reload_insns): Skip debug markers.
3487 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3489         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3490         for its nop_endbr.
3492         PR bootstrap/83396
3493         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3494         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3495         * config/sh/sh.c: Skip debug insns besides notes.
3496         * config/sh/sh.md: Likewise.
3497         * config/sh/sh_treg_combine.cc: Likewise.
3498         * config/sh/sync.md: Likewise.
3500 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3502         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3503         Add item for weak_undefined.
3505 2017-12-14  Richard Biener  <rguenther@suse.de>
3507         PR tree-optimization/67842
3508         * sese.h (bb_in_region): Remove #if 0'ed code.
3510 2017-12-14  Richard Biener  <rguenther@suse.de>
3512         PR tree-optimization/83326
3513         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3514         may_be_zero parameter and handle it by not marking the first
3515         peeled copy as not exiting the loop.
3516         (try_peel_loop): Likewise.
3517         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3518         to handle the case of constant or zero iterations and perform
3519         loop header copying on-the-fly.
3521 2017-12-14  Richard Biener  <rguenther@suse.de>
3523         PR tree-optimization/83418
3524         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3525         Instead of asserting we don't get unfolded comparisons deal with them.
3527 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3529         PR bootstrap/83396
3530         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3531         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3532         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3533         bb.  Simplify.
3535         PR tree-optimization/83198
3536         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3537         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3538         value if arg is a REAL_CST with incompatible type.
3540 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3541             Bin Cheng  <bin.cheng@arm.com>
3543         PR target/81228
3544         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3545         to CCFPEmode.
3546         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3547         LTGT.
3549 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3551         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3552         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3554 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3555             Alan Hayward  <alan.hayward@arm.com>
3556             David Sherwood  <david.sherwood@arm.com>
3558         * poly-int.h: New file.
3559         * poly-int-types.h: Likewise.
3560         * coretypes.h: Include them.
3561         (POLY_INT_CONVERSION): Define.
3562         * target.def (estimated_poly_value): New hook.
3563         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3564         * doc/tm.texi: Regenerate.
3565         * doc/poly-int.texi: New file.
3566         * doc/gccint.texi: Include it.
3567         * doc/rtl.texi: Describe restrictions on subreg modes.
3568         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3569         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3570         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3571         * targhooks.h (default_estimated_poly_value): Declare.
3572         * targhooks.c (default_estimated_poly_value): New function.
3573         * target.h (estimated_poly_value): Likewise.
3574         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3575         (wi::unary_traits): Delete.
3576         (wi::binary_traits::signed_shift_result_type): Define for
3577         offset_int << HOST_WIDE_INT, etc.
3578         (generic_wide_int::operator <<=): Define for all types and use
3579         wi::lshift instead of <<.
3580         (wi::hwi_with_prec): Add a default constructor.
3581         (wi::ints_for): New class.
3582         (operator <<): Define for all wide-int types.
3583         (operator /): New function.
3584         (operator %): Likewise.
3585         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3586         (ASSERT_MAYBE_NE_AT): New macros.
3588 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3589             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3591         PR middle-end/78468
3592         * emit-rtl.c (init_emit): Remove ??? comment.
3593         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3594         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3595         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3596         alignment of 3 virtual registers to BITS_PER_WORD.
3598         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3600 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3602         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3603         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3605 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3607         PR bootstrap/83396
3608         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3610 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3612         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3613         with the exception of debug begin stmt markers appear before
3614         labels.
3616         PR bootstrap/83396
3617         * final.c (rest_of_handle_final): Call variable_tracking_main only
3618         if !flag_var_tracking.
3620 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3621             Jakub Jelinek  <jakub@redhat.com>
3623         PR bootstrap/83396
3624         PR debug/83391
3625         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3626         labels debug stmts that can only appear after labels.
3628 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3630         PR rtl-optimization/82398
3631         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3632         EXPR_USEFULNESS in priority comparison.
3634 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3636         PR rtl-optimization/83393
3637         * combine.c (move_deaths): If reg_stat points to a too new insn in
3638         last_death, do not use it: find the proper insn instead.
3640 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3642         PR target/82190
3643         * config/rs6000/rs6000-string.c (expand_block_compare,
3644         expand_strn_compare): Fix set_mem_size() calls.
3646 2017-12-12  Jeff Law  <law@redhat.com>
3648         PR tree-optimization/83298
3649         PR tree-optimization/83362
3650         PR tree-optimization/83383
3651         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3652         push_value_range a public interface.  Add new argument to
3653         record_ranges_from_stmt.
3654         * gimple-ssa-evrp-analyze.c
3655         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3656         Update comments.  Handle recording temporary equivalences.
3657         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3658         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3659         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3660         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3661         gimple-ssa-evrp-analyze.h.
3662         (record_temporary_equivalences_from_phis): Add new argument.  When
3663         the PHI arg is an SSA_NAME, set the result's range to the range
3664         of the PHI arg.
3665         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3666         from statements too.
3667         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3668         Pass it down to children as needed.
3669         (thread_outgoing_edges): Likewise.
3670         (thread_across_edge): Likewise.   Push/pop range state as needed.
3671         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3673 2017-12-12  Julia Koval  <julia.koval@intel.com>
3675         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3676         (PTA_CANNONLAKE): Remove PTA_CLWB.
3678 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3680         PR target/83332
3681         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3682         (vcondv2div2df): Likewise.
3683         (vconduv2dfv2di): Likewise.
3685 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3687         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3688         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3689         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3690         ATTR_MATHFN_FPROUNDING_ERRNO.
3692 2017-12-12  Richard Biener  <rguenther@suse.de>
3694         PR tree-optimization/83385
3695         * graphite-scop-detection.c (get_order, order): Remove.
3696         (bb_to_rpo): New global.
3697         (cmp_pbbs): Adjust.
3698         (build_scops): Sort pbbs in RPO order.
3700 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3702         * combine.c (simplify_set): Do not transform subregs to zero_extends
3703         if the destination mode is a vector mode.
3705 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3707         PR tree-optimization/83359
3708         * tree-cfg.h (fold_loop_internal_call): Declare.
3709         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3710         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3711         (find_loop_dist_alias): New function.
3712         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3713         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3714         calls.
3716         PR tree-optimization/80631
3717         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3718         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3719         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3720         hardcoding zero as the value if COND_EXPR is never true.  For
3721         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3722         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3723         hardcoding MAX_EXPR as the reduction operation.
3724         (is_nonwrapping_integer_induction): Allow negative step.
3725         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3726         vect_create_epilog_for_reduction, if no value is suitable, don't
3727         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3729 2017-12-12  Richard Biener  <rguenther@suse.de>
3731         PR tree-optimization/81889
3732         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3733         range info from the non-wrapping IV instead of just the range
3734         of the type.
3736 2017-12-12  Julia Koval  <julia.koval@intel.com>
3738         * config.gcc: Add vaesintrin.h.
3739         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3740         New type.
3741         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3742         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3743         New builtins.
3744         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3745         * config/i386/immintrin.h: Include vaesintrin.h.
3746         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3747         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3748         _mm_aesdec_epi128): New intrinsics.
3750 2017-12-12  Julia Koval  <julia.koval@intel.com>
3752         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3753         OPTION_MASK_ISA_VAES_UNSET): New.
3754         (ix86_handle_option): Handle -mvaes.
3755         * config/i386/cpuid.h: Define bit_VAES.
3756         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3757         * config/i386/i386-c.c (__VAES__): New.
3758         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3759         (ix86_valid_target_attribute_inner_p): Ditto.
3760         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3761         * config/i386/i386.opt: Add -mvaes.
3762         * doc/invoke.texi: Ditto.
3764 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3766         * debug.h (gcc_debug_hooks): Add inline_entry.
3767         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3768         * debug.c (do_nothing_debug_hooks): Likewise.
3769         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3770         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3771         (dwarf2_lineno_debug_hooks): Likewise.
3773         * common.opt (gstatement-frontiers): New, setting
3774         debug_nonbind_markers_p.
3775         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3776         * toplev.c (process_options): Autodetect value for debug statement
3777         frontiers option.
3778         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3779         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3781         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3782         markers.  Integrate source bind into debug stmt expand loop.
3783         (pass_expand::execute): Check debug marker limit.  Avoid deep
3784         TER and expand debug locations for debug bind insns only.
3785         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3786         followed by them.
3787         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3788         * final.c (reemit_insn_block_notes): Take current block from
3789         nonbind markers.  Declare note where it's first set.
3790         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3791         begin stmt markers if enabled.
3792         (notice_source_line): Handle nonbind markers.  Fail if their
3793         location is unknown or that of builtins.
3794         (rest_of_handle_final): Convert begin stmt markers to notes if
3795         var-tracking didn't run.
3796         (rest_of_clean_state): Skip begin stmt markers.
3797         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3798         markers.
3799         * function.c (allocate_struct_function): Set begin_stmt_markers.
3800         * function.h (struct function): Add debug_marker_count counter
3801         and debug_nonbind_markers flag.
3802         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3803         * gimple-low.c (lower_function_body): Adjust
3804         debug_nonbind_markers.
3805         (lower_stmt): Drop or skip gimple debug stmts.
3806         (lower_try_catch): Skip debug stmts.
3807         * gimple.c (gimple_build_debug_begin_stmt): New.
3808         (gimple_copy): Increment debug_marker_count if copying one.
3809         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3810         * gimplify.c (rexpr_location): New.
3811         (rexpr_has_location): New.
3812         (warn_switch_unreachable_r): Handle gimple debug stmts.
3813         (shortcut_cond_r): Call expr_location.
3814         (find_goto): New.
3815         (find_goto_label): New.
3816         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3817         find_goto_label.
3818         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3819         expr_location.
3820         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3821         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3822         (LANG_HOOKS_INITIALIZER): ... this.
3823         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3824         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3825         debug insns.
3826         (update_ebb_live_info): Skip debug insn markers.
3827         * lra.c (debug_insn_static_data): Rename to...
3828         (debug_bind_static_data): ... this.
3829         (debug_marker_static_data): New.
3830         (lra_set_insn_recog_data): Select one of the above depending
3831         on debug insn kind.
3832         (lra_update_isn_regno_info): Don't assume debug insns have
3833         freqs.
3834         (push_insns): Skip debug insns.
3835         * lto-streamer-in.c (input_function): Drop debug stmts
3836         depending on active options.  Adjust debug_nonbind_markers.
3837         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3838         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3839         begin stmt marker notes.
3840         (print_insn): Likewise.
3841         * recog.c (extract_insn): Recognize rtl for debug markers.
3842         * rtl.def (DEBUG_MARKER): New.
3843         * tree-inline.c: Include params.h.
3844         (remap_gimple_stmt): Handle nonbind markers.
3845         (maybe_move_debug_stmts_to_successors): Likewise.
3846         (copy_debug_stmt): Likewise.
3847         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3848         markers regardless of no side effects.
3849         (tsi_link_before): Don't update container's side effects when adding
3850         a begin stmt marker.
3851         (tsi_link_after): Likewise.
3852         (expr_first): Skip begin stmt markers.
3853         (expr_last): Likewise.
3854         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3855         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3856         Disregard nonbind markers.
3857         * tree.c (make_node_stat): Don't set side effects for begin stmt
3858         markers.
3859         (build1_stat): Likewise.
3860         * tree.def (DEBUG_BEGIN_STMT): New.
3861         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3862         * var-tracking.c (delete_debug_insns): Renamed to...
3863         (delete_vta_debug_insns): ... this.
3864         (reemit_marker_as_note): New.
3865         (vt_initialize): Reemit markers.
3866         (delete_vta_debug_insns): Likewise.
3867         (vt_debug_insns_local): Reemit or delete markers.
3868         (variable_tracking_main_1): Likewise.
3869         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
3870         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
3871         (gimple_debug_nonbind_marker_p): New.
3872         (gimple_build_debug_bind): Adjust.
3873         (gimple_build_debug_begin_stmt): New.
3874         * doc/invoke.texi (max-debug-marker-count): New param.
3875         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
3876         (debug_parameter_ref, debug_marker): New.
3877         (NOTE_INSN_BEGIN_STMT): New.
3878         (DEBUG_INSN): Describe begin stmt markers.
3880         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
3881         without debug insns, we wouldn't, but clean up debug insns
3882         after a control flow insn nevertheless.
3884         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
3885         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
3886         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
3887         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
3888         (rtl_tidy_fallthru_edge): Likewise.
3889         (rtl_verify_fallthru): Likewise.
3890         (rtl_verify_bb_layout): Likewise.
3891         (skip_insns_after_block): Likewise.
3892         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
3893         * dwarf2out.c: Include print-rtl.h.
3894         (dwarf2out_next_real_insn): New.
3895         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
3896         Dump debug binds in asm comments.
3897         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
3898         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
3899         callers to use gsi_start_nondebug_bb instead.
3900         (gsi_after_labels): Skip gimple debug stmts.
3901         (gsi_start_nondebug): New.
3902         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
3903         (proper_loop_form_for_interchange): Adjust.
3904         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
3905         * gimple.h (gimple_seq_last_nondebug_stmt): New.
3906         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
3907         (collect_fallthrough_labels): Likewise.
3908         (should_warn_for_implicit_fallthrough): Likewise.
3909         (warn_implicit_fallthrough_r): Likewise.
3910         (expand_FALLTHROUGH_r): Likewise.
3911         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
3912         (graphite_copy_stmts_from_block): Skip nonbind markers.
3913         * haifa-sched.c (sched_extend_bb): Skip debug insns.
3914         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
3915         * jump.c (clean_barriers): Skip debug insns.
3916         * omp-expand.c (expand_parallel_call): Skip debug insns.
3917         (expand_task_call): Likewise.
3918         (remove_exit_barrier): Likewise.
3919         (expand_omp_taskreg): Likewise.
3920         (expand_omp_for_init_counts): Likewise.
3921         (expand_omp_for_generic): Likewise.
3922         (expand_omp_for_static_nochunk): Likewise.
3923         (expand_omp_for_static_chunk): Likewise.
3924         (expand_omp_simd): Likewise.
3925         (expand_omp_taskloop_for_outer): Likewise.
3926         (expand_omp_taskloop_for_inner): Likewise.
3927         (expand_oacc_for): Likewise.
3928         (expand_omp_sections): Likewise.
3929         (expand_omp_single): Likewise.
3930         (expand_omp_synch): Likewise.
3931         (expand_omp_atomic_load): Likewise.
3932         (expand_omp_atomic_store): Likewise.
3933         (expand_omp_atomic_fetch_op): Likewise.
3934         (expand_omp_atomic_pipeline): Likewise.
3935         (expand_omp_atomic_mutex): Likewise.
3936         (expand_omp_target): Likewise.
3937         (grid_expand_omp_for_loop): Likewise.
3938         (grid_expand_target_grid_body): Likewise.
3939         (build_omp_regions_1): Likewise.
3940         * omp-low.c (check_combined_parallel): Skip debug stmts.
3941         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
3942         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
3943         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
3944         test.
3945         * tree-cfg.c (make_blobs_1): Skip debug stmts.
3946         (make_edges): Likewise.
3947         (cleanup_dead_labels): Likewise.
3948         (gimple_can_merge_blocks_p): Likewise.
3949         (stmt_starts_bb_p): Likewise.
3950         (gimple_block_label): Likewise.
3951         (gimple_redirect_edge_and_branch): Likewise.
3952         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
3953         of debug stmts.
3954         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
3955         TDF_SLIM.
3956         * tree-pretty-print (print_declaration): Omit initializer in slim
3957         dumps.
3958         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
3959         markers.
3960         (eliminate_unnecessary_stmts): Stabilize block removal order.
3961         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
3962         * var-tracking.c (get_first_insn): New.
3963         (vt_emit_notes): Call it.
3964         (vt_initialize): Walk any insns before the first BB.
3965         (delete_debug_insns): Likewise.
3967         * gimple.h (enum gimple_debug_subcode): Add
3968         GIMPLE_DEBUG_BEGIN_STMT.
3969         (gimple_debug_begin_stmt_p): New.
3970         (gimple_debug_nonbind_marker_p): New.
3971         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
3972         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
3973         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
3974         * insn-notes.def (BEGIN_STMT): New.
3975         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
3976         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
3977         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
3978         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
3979         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
3980         (INSN_DEBUG_MARKER_KIND): New.
3981         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
3982         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
3983         (INSN_VAR_LOCATION_PTR): New.
3984         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
3985         only.
3986         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
3987         deps only if debug bind insns are enabled.
3988         (pass_expand::execute): Avoid deep TER and expand
3989         debug locations for debug bind insns only.
3990         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
3991         debug stmts special handling down to debug bind stmts.
3992         * combine.c (try_combine): Narrow debug insns special handling
3993         down to debug bind insns.
3994         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
3995         Narrow debug insns preexisting special handling down to debug
3996         bind insns.
3997         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
3998         handling down to debug bind insns.
3999         * function.c (instantiate_virtual_regs): Skip debug markers,
4000         adjust handling of debug binds.
4001         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4002         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4003         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4004         insns to debug bind insns.
4005         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4006         special handling of debug stmts to debug bind stmts.
4007         * ipa-split.c (split_function): Likewise.
4008         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4009         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4010         debug insns.
4011         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4012         * regrename.c (build_def_use): Likewise.
4013         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4014         (pass_cprop_hardreg): Narrow special casing of debug insns to
4015         debug bind insns.
4016         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4017         * reload1.c (reload): Likewise.
4018         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4019         casing of debug stmts to debug bind stmts.
4020         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4021         * ssa-iterators.h (num_imm_uses): Likewise.
4022         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4023         debug stmts to debug bind stmts.
4024         * tree-inline.c (tree_function_versioning): Narrow special casing
4025         of debug stmts to debug bind stmts.
4026         * tree-loop-distribution.c (generate_loops_for_partition):
4027         Narrow special casing of debug stmts to debug bind stmts.
4028         * tree-sra.c (analyze_access_subtree): Narrow special casing
4029         of debug stmts to debug bind stmts.
4030         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4031         stmts to debug bind stmts.
4032         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4033         casing of debug stmts to debug bind stmts.
4034         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4035         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4036         casing of debug stmts to debug bind stmts.
4037         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4038         Likewise.
4039         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4040         debug stmts to debug bind stmts.
4041         (gimple_replace_ssa_lhs): Likewise.
4042         (insert_debug_temp_for_var_def): Likewise.
4043         (insert_debug_temps_for_defs): Likewise.
4044         (reset_debug_uses): Likewise.
4045         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4046         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4047         (adjust_debug_stmts): Likewise.
4048         (adjust_phi_and_debug_stmts): Likewise.
4049         (vect_do_peeling): Likewise.
4050         * tree-vect-loop.c (vect_transform_loop): Likewise.
4051         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4052         * var-tracking.c (adjust_mems): Narrow special casing of debug
4053         insns to debug bind insns.
4054         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4055         (compute_bb_dataflow, vt_find_locations): Likewise.
4056         (vt_expand_loc, emit_notes_for_changes): Likewise.
4057         (vt_init_cfa_base): Likewise.
4058         (vt_emit_notes): Likewise.
4059         (vt_initialize): Likewise.
4060         (vt_finalize): Likewise.
4062         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4063         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4064         (next_nonnote_nondebug_insn_bb): New.
4065         (prev_nonnote_nondebug_insn_bb): New.
4066         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4067         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4068         (prev_nonnote_nondebug_insn_bb): Declare.
4069         (next_nonnote_nondebug_insn_bb): Declare.
4070         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4071         * cfgrtl.c (get_last_bb_insn): Likewise.
4072         * lra.c (push_insns): Likewise.
4074 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4076         PR c/82050
4077         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4078         to...
4079         * selftest.c (selftest::test_runner::test_runner): New ctor.
4080         (selftest::test_runner::~test_runner): New dtor.
4081         * selftest.h (class selftest::test_runner): New class.
4083 2017-12-11  Carl Love  <cel@us.ibm.com>
4085         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4086         vec_extract_fp32_from_shortl]): Add #defines.
4087         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4088         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4089         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4090         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4091         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4092         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4093         * doc/extend.texi: Add documentation for the added builtins.
4095 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4097         PR rtl-optimization/80693
4098         PR rtl-optimization/81019
4099         PR rtl-optimization/81020
4100         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4101         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4102         possibly modified to REG_DEAD, if it did not originate in i3.
4104 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4106         * recog.c (store_data_bypass_p_1): New function.
4107         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4108         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4110 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4112         PR rtl-optimization/83361
4113         * ifcvt.c (if_convert): Call fixup_partitions.
4115 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4117         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4118         early folding of splat_u{8,16,32}.
4120 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4122         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4123         output_operand_lossage first argument with capital letter.
4124         (aarch64_override_options): Don't start error and sorry first argument
4125         with capital letter.
4127 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4129         PR gcov-profile/83355
4130         * auto-profile.c (string_table::get_index_by_decl): Don't
4131         recurse when abstract origin points to itself.
4133 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4135         PR tree-optimization/83320
4136         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4137         (prune_datarefs_not_in_loop): Ditto.
4139 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4141         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4143 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4145         PR tree-optimization/83337
4146         * gimple-loop-interchange.cc (compute_access_stride): Handle
4147         bitfield DRs properly.
4149 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4151         PR tree-optimization/83338
4152         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4153         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4154         vector of unsigned integers to vector of signed integers.
4156 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4158         PR rtl-optimization/83317
4159         * lra-constraints.c (process_address_1): Add insn code check.
4161 2017-12-08  Michael Matz  <matz@suse.de>
4163         Fix PR tree-optimization/83323
4164         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4165         head-controlled loops and loop BBs.
4166         * common.opt (funroll-and-jam): Remove, instead ...
4167         (floop-unroll-and-jam): ... reuse this option.
4168         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4169         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4170         (-floop-unroll-and-jam): ... this option.
4172 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4174         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4175         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4176         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4178 2017-12-08  Richard Biener  <rguenther@suse.de>
4180         PR middle-end/81782
4181         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4182         handle accesses outside of zero-sized vars.
4184 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4186         PR tree-optimization/83141
4187         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4188         test for MEM_REFs implicitely changing types with padding.  Remove
4189         inline keyword.
4190         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4192 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4194         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4195         the comment.
4196         * config/avr/avr.c (avr_attribute_table): Likewise.
4197         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4198         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4199         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4200         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4201         of struct attribute_spec.
4202         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4204 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4206         PR target/82960
4207         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4209 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4211         PR middle-end/83609
4212         * profile-count.c (profile_count::from_gcov_type): Move from
4213         profile-count.h; handle overflow.
4214         * profile-count.h (profile_count::from_gcov_type): Move offline.
4216 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4218         PR rtl-optimization/83304
4219         * combine.c (move_deaths): If we do not know where a register died,
4220         search for it.
4222 2017-12-08  Richard Biener  <rguenther@suse.de>
4224         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4225         Provide -fopt-info-loop feedback when we interchange in a nest.
4227 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4229         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4230         for armv6 ARM CPU IDs.
4232 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4234         * common/config/arm/arm-common.c: Include <algorithm>.
4235         (INCLUDE_VECTOR): Define.
4236         (compare_opt_names): New function.
4237         (arm_rewrite_selected_arch): Only strip out extensions that can be
4238         expressed through -mfpu.  Sort the remaining extensions
4239         alphabetically.
4241 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4243         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4244         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4245         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4246         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4248 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4250         PR ada/81470
4251         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4252         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4254 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4256         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4257         of struct attribute_spec.
4259 2017-12-08  Julia Koval  <julia.koval@intel.com>
4261         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4262         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4263         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4264         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4265         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4266         _mm_maskz_dpwssds_epi32): Ditto.
4268 2017-12-08  Richard Biener  <rguenther@suse.de>
4270         PR tree-optimization/81303
4271         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4272         conditions try to create a comparison vector type matching
4273         the data vector type.
4274         (vectorizable_condition): Adjust.
4275         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4276         Leave invariant conditions alone in case we can vectorize those.
4278 2017-12-08  Julia Koval  <julia.koval@intel.com>
4280         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4281         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4282         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4283         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4284         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4286 2017-12-08  Julia Koval  <julia.koval@intel.com>
4288         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4289         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4290         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4291         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4292         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4293         _mm_maskz_dpbusds_epi32): New intrinsics.
4295 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4297         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4298         operand.
4299         (ld<bh>io_signed): Likewise.
4300         (st<bhw>io): Likewise.
4301         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4302         SMALL_INT12 constant integer operand.
4304 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4306         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4307         Rename to...
4308         (nios2_large_constant_allowed): ...this.  Adjust uses.
4309         (nios2_plus_symbolic_constant_p): Rename to...
4310         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4311         (nios2_legitimate_address_p): Correct CONST_INT handling.
4312         (nios2_symbolic_memory_operand_p): Rename to...
4313         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4314         (nios2_large_constant_p): Check for large constant integers too.
4315         (nios2_split_large_constant): Handle constant integers.
4316         (nios2_split_symbolic_memory_operand): Rename to...
4317         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4318         (nios2_legitimize_constant_address): Handle constant integers.
4319         (r0rel_constant_p): Handle small constant integers.
4320         (nios2_print_operand_address): Handle r0-relative integer addresses.
4321         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4322         * config/nios2/nios2.md: Adjust for renamed functions.
4324 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4326         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4327         (TARGET_HAVE_SRODATA_SECTION): New define.
4328         (riscv_select_section): New function.
4330 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4331             Alexander Monakov  <amonakov@ispras.ru>
4332             Jakub Jelinek  <jakub@redhat.com>
4334         PR target/81906
4335         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4337 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4339         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4340         a bool success value.  Don't call output_operand_lossage here.
4341         (aarch64_print_ldpstp_address): Return a bool success value.
4342         (aarch64_print_operand_address): Call output_addr_const if
4343         aarch64_print_address_internal fails.
4344         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4345         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4346         if aarch64_print_ldpstp_address fails.
4348 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4350         * tree-vector-builder.h
4351         (tree_vector_builder::binary_encoded_nelts): Declare.
4352         * tree-vector-builder.c
4353         (tree_vector_builder::binary_encoded_nelts): New function.
4354         * fold-const.c (negate_expr_p): Likewise.
4355         (operand_equal_p, fold_checksum_tree): Likewise.
4356         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4357         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4358         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4359         (uniform_vector_p): Likewise.
4360         * varasm.c (const_hash_1, compare_constant): Likewise.
4361         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4362         (valid_lattice_transition): Operate directly on the VECTOR_CST
4363         encoding.
4364         * ipa-icf.c: Include tree-vector-builder.h.
4365         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4366         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4368 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4370         * tree.c (build_vector): Delete.
4371         * tree.h (build_vector): Make static and move into the self-testing
4372         block.
4374 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4376         * vector-builder.h (vector_builder::derived): New const overload.
4377         (vector_builder::elt): New function.
4378         * tree-vector-builder.h (tree_vector_builder::type): New function.
4379         (tree_vector_builder::apply_step): Declare.
4380         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4381         function.
4382         * gimple-fold.h (tree_vector_builder): Declare.
4383         (gimple_build_vector): Take a tree_vector_builder instead of a
4384         type and vector of elements.
4385         * gimple-fold.c (gimple_build_vector): Likewise.
4386         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4387         accordingly.
4388         (get_initial_defs_for_reduction): Likewise.
4389         (vectorizable_induction): Likewise.
4391 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4393         * tree-vector-builder.h
4394         (tree_vector_builder::new_binary_operation): Declare.
4395         * tree-vector-builder.c
4396         (tree_vector_builder::new_binary_operation): New function.
4397         * fold-const.c (fold_relational_const): Use it.
4398         (const_binop): Likewise.  Check that both input vectors have
4399         the same number of elements, thus excluding things like WIDEN_SUM.
4400         Check whether it is possible to operate directly on the encodings
4401         of stepped inputs.
4403 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4405         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4406         new_unary_operation, operating only on the encoded elements.
4407         (const_unop): Likewise.
4408         (exact_inverse): Likewise.
4409         (distributes_over_addition_p): New function.
4410         (const_binop): Use tree_vector_builder and new_unary_operation
4411         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4412         on the encoded elements unless the encoding is strided and the
4413         operation does not distribute over addition.
4414         (fold_convert_const):  Use tree_vector_builder and
4415         new_unary_operation.  Operate only on the encoded elements
4416         for truncating integer conversions, or for non-stepped encodings.
4418 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4420         * config/sparc/sparc.c: Include tree-vector-builder.h.
4421         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4422         * expmed.c: Include tree-vector-builder.h.
4423         (make_tree): Use tree_vector_builder instead of build_vector.
4424         * fold-const.c: Include tree-vector-builder.h.
4425         (const_binop): Use tree_vector_builder instead of build_vector.
4426         (const_unop): Likewise.
4427         (native_interpret_vector): Likewise.
4428         (fold_vec_perm): Likewise.
4429         (fold_ternary_loc): Likewise.
4430         * gimple-fold.c: Include tree-vector-builder.h.
4431         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4432         of build_vector.
4433         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4434         (simplify_vector_constructor): Use tree_vector_builder instead
4435         of build_vector.
4436         * tree-vect-generic.c: Include tree-vector-builder.h.
4437         (add_rshift): Use tree_vector_builder instead of build_vector.
4438         (expand_vector_divmod): Likewise.
4439         (optimize_vector_constructor): Likewise.
4440         * tree-vect-loop.c: Include tree-vector-builder.h.
4441         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4442         of build_vector.  Explicitly use a stepped encoding for
4443         { 1, 2, 3, ... }.
4444         * tree-vect-slp.c: Include tree-vector-builder.h.
4445         (vect_get_constant_vectors): Use tree_vector_builder instead
4446         of build_vector.
4447         (vect_transform_slp_perm_load): Likewise.
4448         (vect_schedule_slp_instance): Likewise.
4449         * tree-vect-stmts.c: Include tree-vector-builder.h.
4450         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4451         (vect_gen_perm_mask_any): Likewise.
4452         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4453         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4454         of build_vector.
4455         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4456         encoding.
4458 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4460         * doc/generic.texi (VECTOR_CST): Describe new representation of
4461         vector constants.
4462         * vector-builder.h: New file.
4463         * tree-vector-builder.h: Likewise.
4464         * tree-vector-builder.c: Likewise.
4465         * Makefile.in (OBJS): Add tree-vector-builder.o.
4466         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4467         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4468         (tree_vector): Change the number of elements to
4469         vector_cst_encoded_nelts.
4470         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4471         (VECTOR_CST_ELTS): Delete.
4472         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4473         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4474         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4475         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4476         (VECTOR_CST_ENCODED_ELT): Likewise.
4477         (vector_cst_encoded_nelts): New function.
4478         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4479         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4480         (vector_cst_int_elt, vector_cst_elt): Declare.
4481         * tree.c: Include tree-vector-builder.h.
4482         (tree_code_size): Abort if passed VECTOR_CST.
4483         (tree_size): Update for new VECTOR_CST layout.
4484         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4485         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4486         (build_vector): Use tree_vector_builder.
4487         (vector_cst_int_elt, vector_cst_elt): New functions.
4488         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4489         encoded elements and then create the vector in the canonical form.
4490         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4491         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4492         (tree_c_tests): Call test_vector_cst_patterns.
4493         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4494         VECTOR_CST fields.
4495         (hash_tree): Likewise.
4496         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4497         (streamer_write_tree_header): Likewise.
4498         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4499         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4500         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4502 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4504         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4505         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4506         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4508 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4509             Richard Biener  <rguenther@suse.de>
4511         PR tree-optimization/81303
4512         * Makefile.in (gimple-loop-interchange.o): New object file.
4513         * common.opt (floop-interchange): Reuse the option from graphite.
4514         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4515         -floop-interchange and mention it for -O3.
4516         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4517         * gimple-loop-interchange.cc: New file.
4518         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4519         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4520         * passes.def (pass_linterchange): New pass.
4521         * timevar.def (TV_LINTERCHANGE): New time var.
4522         * tree-pass.h (make_pass_linterchange): New declaration.
4523         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4524         interchange.  Record IV before/after increment in new parameters.
4525         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4526         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4527         path check into...
4528         (check_reduction_path): ...New function here.
4529         * tree-vectorizer.h (check_reduction_path): New declaration.
4531 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4533         PR target/83252
4534         PR rtl-optimization/80818
4535         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4536         always early clobbered.
4537         * lra-lives.c (process_bb_lives): Check input hard regs for early
4538         clobbered non-operand hard reg.
4540 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4542         PR middle-end/83164
4543         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4544         types_compatible_p, just that TYPE_MODE is the same.
4546 2017-12-07  Martin Sebor  <msebor@redhat.com>
4548         PR c/81544
4549         * attribs.c (empty_attribute_table): Initialize new member of
4550         struct attribute_spec.
4551         (decl_attributes): Add argument.  Handle mutually exclusive
4552         combinations of attributes.
4553         (selftests::test_attribute_exclusions): New function.
4554         (selftests::attribute_c_tests): Ditto.
4555         * attribs.h (decl_attributes): Add default argument.
4556         * selftest.h (attribute_c_tests): Declare.
4557         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4558         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4559         member.
4560         * doc/extend.texi (Common Function Attributes): Update const and pure.
4561         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4562         of struct attribute_spec.
4563         * config/arc/arc.c (arc_attribute_table): Same.
4564         * config/arm/arm.c (arm_attribute_table): Same.
4565         * config/avr/avr.c ( avr_attribute_table): Same.
4566         * config/bfin/bfin.c (bfin_attribute_table): Same.
4567         * config/cr16/cr16.c (cr16_attribute_table): Same.
4568         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4569         * config/h8300/h8300.c (h8300_attribute_table): Same.
4570         * config/i386/i386.c (ix86_attribute_table): Same.
4571         * config/ia64/ia64.c (ia64_attribute_table): Same.
4572         * config/m32c/m32c.c (m32c_attribute_table): Same.
4573         * config/m32r/m32r.c (m32r_attribute_table): Same.
4574         * config/m68k/m68k.c (m68k_attribute_table): Same.
4575         * config/mcore/mcore.c (mcore_attribute_table): Same.
4576         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4577         * config/mips/mips.c (mips_attribute_table): Same.
4578         * config/msp430/msp430.c (msp430_attribute_table): Same.
4579         * config/nds32/nds32.c (nds32_attribute_table): Same.
4580         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4581         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4582         * config/rl78/rl78.c (rl78__attribute_table): Same.
4583         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4584         * onfig/rx/rx.c (rx_attribute_table): Same.
4585         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4586         * config/sh/sh.c (sh_attribute_table): Same.
4587         * config/sparc/sparc.c (sparc_attribute_table): Same.
4588         * config/spu/spu.c (spu_attribute_table): Same.
4589         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4590         * config/v850/v850.c (v850_attribute_table): Same.
4591         * config/visium/visium.c (visium_attribute_table): Same.
4593 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4595         PR target/82641
4596         * config/arm/arm.c (INCLUDE_STRING): Define.
4597         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4598         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4599         and .fpu.
4601 2017-12-07  Michael Matz  <matz@suse.de>
4603         Add unroll and jam pass
4605         * gimple-loop-jam.c: New file.
4606         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4607         * common.opt (funroll-and-jam): New option.
4608         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4609         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4610         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4611         * passes.def: Add pass_loop_jam.
4612         * timevar.def (TV_LOOP_JAM): Add.
4613         * tree-pass.h (make_pass_loop_jam): Declare.
4614         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4615         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4616         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4617         to flow_loop_tree_node_add.
4618         (duplicate_subloops, copy_loops_to): Append to sibling list.
4619         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4620         * doc/invoke.texi (-funroll-and-jam): Document new option.
4621         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4623 2017-12-07  Richard Biener  <rguenther@suse.de>
4625         PR tree-optimization/83296
4626         PR tree-optimization/67769
4627         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4628         flow sensitive info in an unrelated BB.
4629         (value_replacement): Use reset_flow_sensitive_info.
4630         (minmax_replacement): Reset flow sensitive info on the def
4631         we move.  Do not reset flow sensitive info in the whole BB
4632         we move the stmt to.
4633         (abs_replacement): Likewise.
4635 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4637         PR target/43871
4638         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4639         rs6000_cpu to the given -mcpu=, or to the default processor.
4641 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4643         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4644         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4645         instead of rs6000_cpu_attr.
4646         (rs6000_adjust_cost): Ditto.
4647         (is_microcoded_insn): Ditto.
4648         (rs6000_adjust_priority): Ditto.
4649         (rs6000_issue_rate): Ditto.
4650         (rs6000_use_sched_lookahead): Ditto.
4651         (rs6000_use_sched_lookahead_guard): Ditto.
4652         (rs6000_sched_reorder): Ditto.
4653         (force_new_group): Ditto.
4654         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4655         (group_ending_nop): Ditto.
4657 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4659         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4660         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4661         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4662         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4663         appropriate.
4664         (rs6000_reassociation_width): Ditto.
4665         (rs6000_emit_epilogue): Ditto.
4666         (rs6000_adjust_cost): Ditto.
4667         (is_microcoded_insn): Ditto.
4668         (is_cracked_insn): Ditto.
4669         (rs6000_adjust_priority): Ditto.
4670         (rs6000_sched_reorder): Ditto.
4671         (rs6000_sched_reorder2): Ditto.
4672         (insn_must_be_first_in_group): Ditto.
4673         (insn_must_be_last_in_group): Ditto.
4674         (rs6000_register_move_cost): Ditto.
4675         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4676         rs6000_cpu.
4678 2017-12-07  Julia Koval  <julia.koval@intel.com>
4680         * config.gcc: Add vaesintrin.h.
4681         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4682         New type.
4683         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4684         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4685         New builtins.
4686         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4687         * config/i386/immintrin.h: Include vaesintrin.h.
4688         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4689         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4690         _mm_aesdec_epi128): New intrinsics.
4692 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4694         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4695         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4696         c-family/c-spellcheck.cc.
4697         (best_macro_match::best_macro_match): Likewise.
4698         * spellcheck-tree.h
4699         (struct edit_distance_traits<cpp_hashnode *>): Move to
4700         c-family/c-spellcheck.h.
4701         (class best_macro_match): Likewise.
4703 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4705         PR tree-optimization/83293
4706         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4707         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4708         might insert into empty bb.
4710         PR sanitizer/81281
4711         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4712         simplify for plus with :c added, and pointer_plus without that.
4713         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4714         with undefined overflow and the conversion is not widening,
4715         perform negation in utype and only convert to type afterwards.
4716         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4717         simplify for plus with :c added, and pointer_plus without that.
4718         If type is integral with undefined overflow and the conversion is
4719         not widening, perform minus in utype and only convert to type
4720         afterwards.  Move the last pointer_diff_expr simplify into the
4721         two outermost ifs.
4723 2017-12-06  Martin Sebor  <msebor@redhat.com>
4725         PR tree-optimization/82646
4726         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4727         strncpy, not maxlen.
4729 2017-12-06  Martin Sebor  <msebor@redhat.com>
4731         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4732         nonstring.
4734         PR tree-optimization/83075
4735         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4736         strncat/strncpy don't change length of source string.
4738 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4740         Revert
4741         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4743         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4744         to prevent b2bst errata sequence.
4745         (sqrtdf2_fix): Likewise.
4747 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4749         PR tree-optimization/81945
4750         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4751         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4752         to dest_cfun has orig_loop_num set, either remap it to the new
4753         loop number if the loop got moved too, or clear it.
4755 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4757         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4758         to reservation.
4759         (thunderx2t99_nothing): New insn reservation.
4760         (thunderx2t99_mrs): New insn reservation.
4761         (thunderx2t99_multiple): New insn reservation.
4762         (thunderx2t99_alu_basi): Add bfx to reservation.
4763         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4765 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4767         PR target/82248
4768         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4770 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4772         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4773         tree-ssa-pre.c::remove_dead_inserted_code.
4774         * tree-ssa-dce.h: New file.
4775         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4776         (remove_dead_inserted_code): Move and rename to function
4777         tree-ssa-dce.c::simple_dce_from_worklist.
4778         (pass_pre::execute): Update use.
4780 2017-12-05  Richard Biener  <rguenther@suse.de>
4782         PR tree-optimization/83277
4783         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4784         to code-gen liveout vars.
4786 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4788         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4789         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4790         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4791         split condition.
4793 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4795         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4796         function.
4797         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4798         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4799         ASAN is enabled.
4801 2017-12-05  Richard Biener  <rguenther@suse.de>
4803         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4804         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4805         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4806         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4808 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4810         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4812 2017-12-05  Martin Liska  <mliska@suse.cz>
4813             Jakub Jelinek  <jakub@redhat.com>
4815         * doc/invoke.texi: Document the options.
4816         * flag-types.h (enum sanitize_code): Add
4817         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4818         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4819         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4820         * opts.c: Define new sanitizer options.
4821         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4822         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4824 2017-12-05  Julia Koval  <julia.koval@intel.com>
4826         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4827         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4828         (ix86_handle_option): Handle -mavx512vnni.
4829         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4830         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4831         * config/i386/i386-c (__AVX512VNNI__): New.
4832         * config/i386/i386.c (ix86_target_string): Handle new option.
4833         (ix86_valid_target_attribute_inner_p): Handle new option.
4834         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4835         * config/i386/i386.opt (mavx512vnni): New option.
4837 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4839         PR target/81616
4840         * athlon.md: Disable for generic.
4841         * haswell.md: Enable for generic.
4842         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4843         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4844         to 4.
4845         (ix86_adjust_cost): Move generic to haswell path.
4847 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
4849         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4850         instead of MEM_P in a couple more places.  Fix formatting issues.
4852 2017-12-04  Jim Wilson  <jimw@sifive.com>
4854         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4855         instead of GP_REG_LAST-1.
4856         (riscv_adjust_libcall_cfi_prologue): Likewise.
4857         (riscv_adjust_libcall_cri_epilogue): Likewise.
4858         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4859         comment.
4861 2017-12-04  Luis Machado  <luis.machado@linaro.org>
4863         * ipa-pure-const.c (check_decl): Add missing newline.
4864         (state_from_flags): Likewise.
4866 2017-12-04  Jeff Law  <law@redhat.com>
4868         PR tree-optimizatin/78496
4869         * gimple-ssa-evrp-analyze.h
4870         (evrp_range_analyzer::get_vr_values): Simplify.
4871         * gimple-ssa-evrp-analyze.c: Corresponding changes.
4872         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
4873         and gimple-ssa-evrp-analyze.h.
4874         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
4875         (simplify_stmt_for_jump_threading): Copy a blob of code from
4876         tree-vrp.c to use ranges to simplify statements.
4877         (dom_opt_dom_walker::before_dom_children): Call
4878         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
4879         (dom_opt_dom_walker::after_dom_children): Similarly for
4880         evrp_range_analyzer::leave.
4881         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
4882         conditionals.
4884         * gimple-ssa-evrp-analyze.c
4885         (evrp_range_analyzer::extract_range_from_stmt):  Always use
4886         vr_values::update_value_range so preexisting range info is
4887         medged with new range info, even if the new range is VR_VARYING.
4889 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4891         * combine.c: Adjust comment.
4892         (use_crosses_set_p): Delete.
4893         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
4894         (try_combine): Ditto.
4896 2017-12-04  Richard Biener  <rguenther@suse.de>
4898         PR tree-optimization/83255
4899         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
4900         Re-add zero-iteration check.
4902 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4904         PR rtl-optimization/83245
4905         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
4906         hard registers as earlyclobber, also if not in an asm.
4908 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4910         PR bootstrap/83265
4911         Revert
4912         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4914         PR target/43871
4915         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4916         rs6000_cpu based on cpu_index, not tune_index.
4918 2017-12-04  Richard Biener  <rguenther@suse.de>
4920         PR tree-optimization/83238
4921         * graphite-scop-detection.c (scop_detection::merge_sese): Make
4922         code match comment, rejecting invalid SESE regions.
4924 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
4926         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
4927         require base operand is a REG_POINTER prior to reload on targets
4928         with non-equivalent space registers.
4930 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4932         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
4933         (update_specialized_profile): Fix updating of counts.
4934         (perhaps_add_new_callers): Likewise.
4936 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4938         PR target/81616
4939         * x86-tune.def: Remove obsolette FIXMEs.
4940         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
4941         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
4942         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
4943         Enable for generic.
4944         (X86_TUNE_PAD_RETURNS): Disable for generic.
4946 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
4948         PR tree-optimization/83170
4949         PR tree-optimization/83241
4950         * gimple-ssa-store-merging.c
4951         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
4952         gimple_vuse (ins_stmt) in case it has changed.
4953         (imm_store_chain_info::output_merged_store): Likewise.
4955         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
4956         POINTER_DIFF_EXPR.
4958         PR c++/81212
4959         * tree-cfg.c (pass_warn_function_return::execute): Handle
4960         __builtin_ubsan_handle_missing_return like __builtin_unreachable
4961         with BUILTINS_LOCATION.
4963         PR target/78643
4964         PR target/80583
4965         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
4966         is BLKmode for vector field with vector raw mode, use TYPE_MODE
4967         instead of DECL_MODE.
4969         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
4970         last argument to rtx pointer.
4971         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
4972         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
4973         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
4974         depending on the chosen ISAs.
4975         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
4976         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
4977         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
4978         callers.
4979         * config/i386/sse.md (mov<mode>_internal): Likewise.
4980         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
4982 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4984         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
4986 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4988         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
4989         parameters from prototype.
4990         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
4991         parameters.  Don't print a comment.
4992         (emit_fusion_gpr_load): Adjust.
4993         (emit_fusion_load_store): Adjust.
4994         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
4995         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
4996         comment on the second line.
4998 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5000         PR target/43871
5001         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5002         rs6000_cpu based on cpu_index, not tune_index.
5004 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5006         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5007         which_alternative instead of which_alternative + 1.
5008         (output_asm_insn): Print an extra tab if the template is short.
5010 2017-12-01  Jim Wilson  <jimw@sifive.com>
5012         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5013         comment.
5014         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5015         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5016         reference.
5017         * doc/tm.texi: Regenerate.
5019 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5021         PR target/81959
5022         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5023         whether we can allocate pseudos before trying to fix an address.
5024         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5025         memory address is indexed or indirect.
5026         (floatuns_<mode>si2_hw2): Likewise.
5028 2017-12-01  Jason Merrill  <jason@redhat.com>
5030         * Makefile.in (TAGS): Add c-family/*.cc.
5032 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5034         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5035         (call_value_insn): Likewise.
5036         (sibcall_insn): Likewise.
5037         (sibcall_value_insn): Likewise.
5038         (movsi_aarch64): Likewise.
5039         (movdi_aarch64): Likewise.
5040         (add_losym_): Likewise.
5041         (ldr_got_small_): Likewise.
5042         (ldr_got_small_sidi): Likewise.
5043         (ldr_got_small_28k_): Likewise.
5044         (ldr_got_small_28k_sidi): Likewise.
5045         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5046         Move output_addr_const to symbolic case. Add error check.
5048 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5050         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5051         * config/s390/s390.md: Disable MVC merging peephole if it would
5052         disable operand forwarding.
5053         (new peephole2): Split MVCs if it would turn them into up to 2
5054         forwardable MVCs.
5056 2017-12-01  Richard Biener  <rguenther@suse.de>
5058         PR tree-optimization/83232
5059         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5060         detection of same access. Instead of breaking the group here
5061         do not consider the duplicate.  Add comment explaining real fix.
5063 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5065         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5067 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5069         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5071 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5073         * function.h (struct function): Remove cilk_frame_decl,
5074         is_cilk_function and calls_cilk_spawn fields.
5075         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5076         field.
5077         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5078         cilk_elemental field.
5079         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5080         * target.def: Adjust comment.
5081         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5082         Don't test cilk_elemental.
5084         PR tree-optimization/83233
5085         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5086         bswap_stat name for the struct.
5088         PR c/79153
5089         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5090         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5091         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5092         SWITCH_BREAK_LABEL_P set on the label.
5093         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5094         added for default case if it was missing and not all cases covered.
5095         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5096         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5097         set on the label.
5098         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5099         to avoid -Wimplicit-fallthrough warning.
5100         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5101         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5103         PR tree-optimization/83221
5104         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5105         down by 16.
5106         (init_reassoc): Formatting fix.
5108         PR sanitizer/81275
5109         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5110         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5111         is diagnosed.
5113         PR sanitizer/83219
5114         * tree-cfg.c: Include asan.h.
5115         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5117 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5119         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5121 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5123         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5124         vpcmpeqd instruction.
5126 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5128         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5129         registers type for 128bit mode.
5131 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5133         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5134         of ;;.
5135         * gengtype-state.c (read_state_pair): Likewise.
5136         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5137         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5138         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5139         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5140         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5141         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5142         * builtins.c (fold_builtin_3): Likewise.
5143         * graphite-scop-detection.c
5144         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5145         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5146         Likewise.
5148 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5150         PR sanitizer/81697
5151         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5152         parameter. Return true if ignore_decl_rtl_set_p is true and other
5153         conditions are satisfied.
5154         * asan.h (asan_protect_global): Add new parameter.
5155         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5156         to asan_protect_global calls.
5158 2017-11-30  Jim Wilson  <jimw@sifive.com>
5160         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5161         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5162         -mstrict-align, add info on default value.  Delete redundant lines for
5163         -mabi.  Add missing -mexplicit-relocs docs.
5165 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5167         * config/arc/arc.md (trap): New pattern.
5169 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5171         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5172         instruction to end into a delay slot.
5173         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5174         can be placed into a delay slot against reg_note.
5176 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5178         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5179         labels number of usages.
5181 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5183         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5184         function.
5185         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5187 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5189         PR target/83210
5190         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5191         multiplication by power of 2 constant into two shifts + comparison.
5193 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5195         PR target/81616
5196         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5198 2017-11-30  Richard Biener  <rguenther@suse.de>
5200         PR tree-optimization/83202
5201         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5202         allow_peel argument and guard peeling.
5203         (canonicalize_loop_induction_variables): Likewise.
5204         (canonicalize_induction_variables): Pass false.
5205         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5206         peeling from cunrolli.
5208 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5210         * combine.c (try_combine): Print a message to dump file whenever
5211         I0, I1, or I2 cannot be combined into I3.
5213 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5215         PR rtl-optimization/83156
5216         PR rtl-optimization/82621
5217         * combine.c (try_combine): Don't split an I2 if one of the dests is
5218         set again before I3.  Allow unused dests.
5220 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5222         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5224 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5226         PR rtl-optimization/80818
5227         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5228         recursively.  Use insn code for clobber.
5229         (lra_set_insn_recog_data): Pass the new arg to
5230         collect_non_operand_hard_regs.
5231         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5232         code for clobber.
5233         (lra_update_insn_regno_info): Pass insn to
5234         add_regs_to_insn_regno_info.
5236 2017-11-29  Jim Wilson  <jimw@sifive.com>
5237             Andrew Waterman  <andrew@sifive.com>
5239         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5240         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5241         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5242         use SINGLE_SHIFT_COST.
5243         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5244         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5246 2017-11-29  Julia Koval  <julia.koval@intel.com>
5248         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5249         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5250         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5251         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5252         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5253         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5254         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5255         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5256         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5257         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5258         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5259         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5260         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5261         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5262         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5263         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5264         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5265         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5266         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5267         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5268         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5269         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5270         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5271         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5272         __builtin_ia32_vpshldv_v2di_mask,
5273         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5274         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5275         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5277 2017-11-29  Julia Koval  <julia.koval@intel.com>
5279         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5280         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5281         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5282         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5283         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5284         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5285         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5286         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5287         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5288         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5289         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5290         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5291         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5292         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5293         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5294         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5295         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5296         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5297         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5298         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5299         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5301 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5303         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5304         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5305         workaround.
5307 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5309         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5310         to prevent b2bst errata sequence.
5311         (sqrtdf2_fix): Likewise.
5313 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5315         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5316         (div_sqrt_insn_p): New function.
5317         (sparc_do_work_around_errata): Insert NOP instructions to
5318         prevent sequences that could trigger the TN-0013 errata for
5319         certain LEON3 processors.
5320         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5321         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5322         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5323         (in_branch_delay): Prevent div and sqrt in delay slot if
5324         fix_lost_divsqrt.
5325         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5327 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5329         * config/sparc/sparc.c (atomic_insn_p): New function.
5330         (sparc_do_work_around_errata): Insert NOP instructions to
5331         prevent sequences that could trigger the TN-0010 errata for
5332         UT700.
5333         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5334         instruction referable in atomic_insns_p.
5336 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5338         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5339         (atomic_compare_and_swap_leon3_1): Likewise.
5340         (ldstub): Likewise.
5342 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5344         * config/sparc/sparc.c (fpop_insn_p): New function.
5345         (sparc_do_work_around_errata): Insert NOP instructions to
5346         prevent sequences that could trigger the TN-0012 errata for
5347         GR712RC.
5348         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5349         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5350         (in_branch_annul_delay): Prevent floating-point instructions
5351         in delay slot of annulled integer branch.
5353 2017-11-29  Richard Biener  <rguenther@suse.de>
5355         PR tree-optimization/83202
5356         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5357         (bst_fail): Use it.
5358         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5359         nodes vectorized to the same stmts multiple times.
5360         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5361         (vect_analyze_slp_instance): Adjust.
5362         (scalar_stmts_to_slp_tree_map_t): New typedef.
5363         (vect_schedule_slp_instance): Add a map recording the SLP node
5364         representing the vectorized stmts for a set of scalar stmts.
5365         Avoid code-generating redundancies.
5366         (vect_schedule_slp): Allocate map and pass it down.
5368 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5370         PR c++/83187
5371         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5372         type if it is new.
5374 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5376         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5377         cases for printing LDP/STP memory addresses.
5378         (aarch64_print_address_internal): Renamed from
5379         aarch64_print_operand_address, added parameter, add Pmode check.
5380         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5381         (aarch64_print_operand_address): Indirect to
5382         aarch64_print_address_internal.
5383         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5384         'y' operand output specifier.
5386 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5388         PR middle-end/83185
5389         * tree.c (build_simple_mem_ref_loc): Handle
5390         get_addr_base_and_unit_offset returning a MEM_REF.
5392         PR middle-end/80929
5393         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5395         PR target/80819
5396         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5397         alternative.
5399 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5401         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5402         m_x_offset.
5403         (layout::move_to_column): Likewise.
5405 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5407         PR sanitizer/81275
5408         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5409         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5411 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5412             Martin Jambor  <mjambor@suse.cz>
5414         PR ipa/82808
5415         * tree.h (expr_type_first_operand_type_p): Declare
5416         * tree.c (expr_type_first_operand_type_p): New function.
5417         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5418         (ipa_value_from_jfunc): Adjust declaration.
5419         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5420         Use it as result type for arithmetics, unless it is NULL in which case
5421         be more conservative.
5422         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5423         ipa_get_jf_pass_through_result.
5424         (propagate_vals_across_pass_through): Likewise.
5425         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5426         is to propagate_vals_across_pass_through.
5427         (propagate_constants_across_call): Pass PARM_TYPE to
5428         propagate_scalar_across_jump_function.
5429         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5430         ipa_value_from_jfunc.
5431         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5432         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5433         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5434         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5435         target_type, pass it to ipa_value_from_jfunc.
5436         (update_indirect_edges_after_inlining): Pass parameter type to
5437         try_make_edge_direct_simple_call.
5439 2017-11-28  Jeff Law  <law@redhat.com>
5441         * gimple-ssa-evrp-analyze.c
5442         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5443         refine ranges if scev_initialized_p returns true.
5444         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5446 2017-11-28  Julia Koval  <julia.koval@intel.com>
5448         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5449         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5450         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5451         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5452         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5453         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5454         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5455         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5456         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5457         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5458         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5459         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5460         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5461         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5462         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5463         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5464         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5465         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5466         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5467         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5468         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5469         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5471 2017-11-28  Julia Koval  <julia.koval@intel.com>
5473         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5474         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5475         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5476         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5477         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5478         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5479         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5480         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5481         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5482         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5483         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5484         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5485         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5486         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5487         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5488         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5489         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5490         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5491         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5492         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5493         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5494         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5495         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5496         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5497         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5498         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5499         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5500         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5501         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5502         __builtin_ia32_vpshld_v2di_mask): New builtins.
5503         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5504         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5506 2017-11-28  Richard Biener  <rguenther@suse.de>
5508         PR tree-optimization/80776
5509         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5510         Declare.
5511         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5512         New function.
5513         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5514         If the incoming edge is an effective fallthru because the other
5515         edge only reaches a __builtin_unreachable () then record ranges
5516         derived from the controlling condition in SSA info.
5517         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5518         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5520 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5522         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5523         on mingw build hosts.
5525 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5527         PR debug/81307
5528         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5530 2017-11-28  Richard Biener  <rguenther@suse.de>
5532         PR middle-end/83141
5533         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5534         copies generated from memcpy use a character array as reference
5535         type.
5537 2017-11-28  Julia Koval  <julia.koval@intel.com>
5538             Sebastian Peryt  <sebastian.peryt@intel.com>
5540         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5541         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5542         cilk-common.o, cilk.h, cilk-common.c): Remove.
5543         * builtin-types.def
5544         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5545         * builtins.c (is_builtin_name): Remove cilkplus condition.
5546         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5547         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5548         cilk-builtins.def, cilkplus.def): Remove.
5549         * cif-code.def (CILK_SPAWN): Remove.
5550         * cilk-builtins.def: Delete.
5551         * cilk-common.c: Ditto.
5552         * cilk.h: Ditto.
5553         * cilkplus.def: Ditto.
5554         * config/darwin.h (fcilkplus): Delete.
5555         * cppbuiltin.c: Ditto.
5556         * doc/extend.texi: Remove cilkplus doc.
5557         * doc/generic.texi: Ditto.
5558         * doc/invoke.texi: Ditto.
5559         * doc/passes.texi: Ditto.
5560         * gcc.c (fcilkplus): Remove.
5561         * gengtype.c (cilk.h): Remove.
5562         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5563         support.
5564         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5565         Remove.
5566         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5567         gimplify_call_expr,
5568         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5569         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5570         cilkplus conditions.
5571         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5572         inline_read_section): Ditto.
5573         * ipa-inline-analysis.c (cilk.h): Remove.
5574         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5575         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5576         append_linker_options): Remove condition for fcilkplus.
5577         * lto/lto-lang.c (cilk.h): Remove.
5578         (lto_init): Remove condition for fcilkplus.
5579         * omp-expand.c (expand_cilk_for_call): Delete.
5580         (expand_omp_taskreg, expand_omp_for_static_chunk,
5581         expand_omp_for): Remove cilkplus
5582         conditions.
5583         (expand_cilk_for): Delete.
5584         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5585         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5586         execute_lower_omp, diagnose_sb_0): Ditto.
5587         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5588         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5589         * tree-nested.c: Ditto.
5590         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5591         (dump_generic_node): Ditto.
5592         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5593         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5594         Delete.
5595         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5597 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5599         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5600         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5601         for vector mode and !TARGET_SIMD.
5603 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5605         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5606         Adjust comment.
5607         * tree.h (SWITCH_LABELS): Remove.
5608         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5609         assert SWITCH_BODY is non-NULL.
5610         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5611         handling.
5612         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5614         PR tree-optimization/80788
5615         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5616         has TREE_OVERFLOW set, call drop_tree_overflow.
5618 2017-11-28  Richard Biener  <rguenther@suse.de>
5620         PR tree-optimization/83158
5621         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5623 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5625         PR 81288/target
5626         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5627         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5629 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5631         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5632         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5634 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5636         PR middle_end/82333
5637         * varasm.c (compare_constant): Take the mode of the constants into
5638         account when comparing floating point constants.
5640 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5642         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5643         from explicit instantiation of debug_helper.
5644         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5646 2017-11-27  Richard Biener  <rguenther@suse.de>
5648         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5649         refactor a bit.
5651 2017-11-27  Richard Biener  <rguenther@suse.de>
5653         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5654         (type_hash_canon): Also clear int_cst_hash_table entry for
5655         TYPE_MIN/MAX_VALUE.
5656         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5658 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5660         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5661         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5662         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5664 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5666         * hash-map.h (gt_cleare_cache): Avoid UB.
5668 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5670         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5671         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5672         (decide_unroll_runtime_iterations): Likewise.
5673         (decide_unroll_stupid): Likewise.
5675 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5677         PR target/83109
5678         * config/i386/i386.md: Add a loop with incssp.
5680 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5682         PR tree-optimization/81248
5683         * tree-sra.c (splice_param_accesses): Remove size check.
5684         (decide_one_param_reduction): Fix size check.
5685         * gimple-pretty-print.c (dump_profile): Silence warning.
5686         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5688 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5690         PR debug/81307
5691         * dbxout.c (lastlineno): New variable.
5692         (dbx_debug_hooks): Use dbxout_switch_text_section as
5693         switch_text_section debug hook.
5694         (dbxout_function_end): Switch to current_function_section
5695         rather than function_section.  If crtl->has_bb_partition,
5696         output just one N_FUN, depending on in_cold_section_p.
5697         (dbxout_source_line): Remember last lineno in lastlineno.
5698         (dbxout_switch_text_section): New function.
5699         (dbxout_function_decl): Adjust dbxout_block caller.
5700         (dbx_block_with_cold_children): New function.
5701         (dbxout_block): Return true if any LBRAC/RBRAC have been
5702         emitted.  Use dbx_block_with_cold_children at depth == 0
5703         in second partition.  Add PARENT_BLOCKNUM argument, pass
5704         it optionally adjusted to children.  Output LBRAC/RBRAC
5705         around recursive call only if the block is in the current
5706         partition, if not and anything was output, emit empty
5707         range LBRAC/RBRAC.
5708         * final.c (final_scan_insn): Compute cold_function_name
5709         before calling switch_text_section debug hook.  Call
5710         that hook even if dwarf2out_do_frame if not emitting
5711         dwarf debug info.
5713         PR target/83100
5714         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5715         TREE_READONLY decls.
5717 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5719         PR rtl-optimization/82488
5720         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5722 2017-11-26  Julia Koval  <julia.koval@intel.com>
5724         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5725         skylake-avx512.
5726         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5727         skylake_cost): New.
5729 2017-11-26  Julia Koval  <julia.koval@intel.com>
5731         * config/i386/driver-i386.c (host_detect_local_cpu):
5732         Detect skylake-avx512.
5734 2017-11-26  Julia Koval  <julia.koval@intel.com>
5736         * config.gcc: Add -march=cannonlake.
5737         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5738         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5739         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5740         (PTA_CANNONLAKE): New.
5741         (processor_target_table): Add cannonlake.
5742         (ix86_option_override_internal): Ditto.
5743         (fold_builtin_cpu): Ditto.
5744         (get_builtin_code_for_version): Handle cannonlake.
5745         (M_INTEL_COREI7_CANNONLAKE): New.
5746         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5747         * doc/invoke.texi: Add -march=cannonlake.
5749 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5751         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5752         (try_init_one_plugin): Alternative implementation for MinGW.
5753         * Makefile.in (plugin_implib): New.
5754         (gengtype-lex.c): Fix broken AIX workaround.
5755         * configure: Regenerate.
5756         * doc/plugins.texi: Document support for MinGW.
5758 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5760         PR rtl-optimization/81553
5761         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5762         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5763         is a shift where C1 has different mode than the whole shift, use C1's
5764         mode for MULT rather than the shift's mode.
5766         PR target/82848
5767         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5768         builtins not enabled in the currently selected ISA.
5770 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5772         PR tree-optimization/71026
5773         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5774         (insert_reciprocals): Change to insert reciprocals before a division
5775         by a square and to insert the square of a reciprocal.
5776         (execute_cse_reciprocals_1): Change to consider division by a square.
5777         (register_division_in): Add importance parameter.
5779 2017-11-24  Richard Biener  <rguenther@suse.de>
5781         PR tree-optimization/82402
5782         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5783         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5785 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5787         * match.pd (0-ptr): New transformation.
5789 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5791         PR bootstrap/83015
5792         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5793         when skipping checking.
5795 2017-11-24  Richard Biener  <rguenther@suse.de>
5797         PR tree-optimization/83128
5798         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5799         (vn_reference_lookup_3): Likewise.
5801 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5803         PR sanitizer/83014
5804         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5805         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5806         tree_to_uhwi twice.
5808         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5809         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5810         check for the latter with gimple_call_builtin_p.  Do not handle
5811         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5813 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5815         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5816         vdot_u32.
5818 2017-11-23  Julia Koval  <julia.koval@intel.com>
5820         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5821         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5822         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5823         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5824         _mm512_maskz_expandloadu_epi16): New intrinsics.
5825         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5826         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5827         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5828         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5829         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5830         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5831         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5832         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5833         _mm256_maskz_expandloadu_epi8): New intrinsics.
5834         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5835         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5836         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5837         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5838         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5839         * config/i386/sse.md (VI248_VLBW): New iterator.
5840         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5842 2017-11-23  Julia Koval  <julia.koval@intel.com>
5844         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5845         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5846         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5847         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5848         _mm512_mask_compressstoreu_epi16): New.
5849         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5850         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5851         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5852         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5853         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5854         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5855         _mm256_mask_compressstoreu_epi8): New.
5856         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5857         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5858         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5859         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5860         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5861         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5862         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
5863         __builtin_ia32_compresshi128_mask,
5864         __builtin_ia32_compressstoreuqi512_mask,
5865         __builtin_ia32_compressstoreuhi512_mask,
5866         __builtin_ia32_compressstoreuqi256_mask,
5867         __builtin_ia32_compressstoreuqi128_mask,
5868         __builtin_ia32_compressstoreuhi256_mask,
5869         __builtin_ia32_compressstoreuhi128_mask): New builtins.
5870         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
5871         array for flags2.
5872         (ix86_expand_special_args_builtin): Handle new types.
5873         (s4fma_expand): Handle new builtin array.
5874         * config/i386/immintrin.h: Include new headers.
5875         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
5876         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
5878 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
5880         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
5881         fixes.  Declare temp and g variables at the top in order to avoid
5882         {} in most of the cases.
5884 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
5886         * match.pd (ptr-0): New transformation.
5888 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5890         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
5891         (struct addr_mode_cost_table): New.
5892         (struct tune_params): Add field addr_mode_costs.
5893         * config/arm/arm.c (generic_addr_mode_costs): New.
5894         (arm_slowmul_tune): Initialise addr_mode_costs field.
5895         (arm_fastmul_tune): Likewise.
5896         (arm_strongarm_tune): Likewise.
5897         (arm_xscale_tune): Likewise.
5898         (arm_9e_tune): Likewise.
5899         (arm_marvell_pj4_tune): Likewise.
5900         (arm_v6t2_tune): Likewise.
5901         (arm_cortex_tune): Likewise.
5902         (arm_cortex_a8_tune): Likewise.
5903         (arm_cortex_a7_tune): Likewise.
5904         (arm_cortex_a15_tune): Likewise.
5905         (arm_cortex_a35_tune): Likewise.
5906         (arm_cortex_a53_tune): Likewise.
5907         (arm_cortex_a57_tune): Likewise.
5908         (arm_exynosm1_tune): Likewise.
5909         (arm_xgene1_tune): Likewise.
5910         (arm_cortex_a5_tune): Likewise.
5911         (arm_cortex_a9_tune): Likewise.
5912         (arm_cortex_a12_tune): Likewise.
5913         (arm_cortex_a73_tune): Likewise.
5914         (arm_v7m_tune): Likewise.
5915         (arm_cortex_m7_tune): Likewise.
5916         (arm_v6m_tune): Likewise.
5917         (arm_fa726te_tune): Likewise.
5918         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
5920 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5922         * config/arm/arm.c (arm_mem_costs): New function.
5923         (arm_rtx_costs_internal): Use arm_mem_costs.
5925 2017-11-23  Mark Wielaard  <mark@klomp.org>
5927         * dwarf2out.c (init_sections_and_labels): Use generation to create
5928         unique ranges_section_label and ranges_base_label. Return generation.
5929         (output_rnglists): Add generation argument. Use generation to create
5930         unique ranges labels.
5931         (dwarf2out_finish): Get generation from init_sections_and_labels
5932         and pass generation to output_rnglists.
5934 2017-11-23  Mike Stump  <mikestump@comcast.net>
5935             Eric Botcazou  <ebotcazou@adacore.com>
5937         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
5938         * cfgloop.h (struct loop): Add unroll field.
5939         * function.h (struct function): Add has_unroll bitfield.
5940         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
5941         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
5942         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
5943         (pass_rtl_unroll_loops::gate): Likewise.
5944         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
5945         for which loop->unroll==1.
5946         (decide_unroll_constant_iterations): Use note for consistency and
5947         take loop->unroll into account.  Return early if loop->unroll is set.
5948         Fix thinko in existing test.
5949         (decide_unroll_runtime_iterations): Use note for consistency and
5950         take loop->unroll into account.
5951         (decide_unroll_stupid): Likewise.
5952         * lto-streamer-in.c (input_cfg): Read loop->unroll.
5953         * lto-streamer-out.c (output_cfg): Write loop->unroll.
5954         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
5955         New case.
5956         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
5957         (print_loop): Print loop->unroll if set.
5958         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
5959         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
5960         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
5961         New case.
5962         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
5963         loop->unroll is set and smaller than the trip count.  Otherwise bypass
5964         entirely the heuristics if loop->unroll is set.  Remove dead note.
5965         Fix off-by-one bug in other note.
5966         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
5967         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
5968         is greater than 1.
5969         (tree_unroll_loops_completely): Make static.
5970         (pass_complete_unroll::execute): Use correct type for variable.
5971         (pass_complete_unrolli::execute): Fix formatting.
5972         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
5974 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5976         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
5977         enable when TARGET_PREFER_AVX128 is set.
5979 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5981         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
5982         frequency.
5983         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
5984         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
5986 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5988         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
5989         frequencies. Fix estimation of aggregate parameters.
5991 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5993         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
5994         when inlining.
5996 2017-11-23  Tom de Vries  <tom@codesourcery.com>
5998         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6000 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6002         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6003         macro.
6005 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6007         PR target/83111
6008         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6009         sibcall_value_pcrel_fdpic): Use local variable instead of
6010         operands[3].
6011         (calli_tbr_rel): Add missing operand 2.
6012         (call_valuei_tbr_rel): Add missing operand 3.
6014 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6016         PR middle-end/82253
6017         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6018         bitpos/bitsize covering the whole destination, use store_expr only if
6019         the complex mode is the same.  Otherwise, use expand_normal and if
6020         it returns CONCAT, subreg each part separately instead of trying to
6021         subreg the whole result.
6023 2017-11-23  Richard Biener  <rguenther@suse.de>
6025         PR tree-optimization/23094
6026         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6027         come here from walking over backedges in the first iteration.
6028         (vn_reference_lookup_3): Skip clobbers that store the same value.
6030 2017-11-23  Richard Biener  <rguenther@suse.de>
6032         PR tree-optimization/81403
6033         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6034         a block we need a leader relative to.
6035         (phi_translate_1): For nary processing require a leader from
6036         get_representative_for given we run expression simplification
6037         using match-and-simplify.  Remove previous fix.
6039 2017-11-22  Jeff Law  <law@redhat.com>
6041         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6042         Use new method allocate_value_range rather than accessing the
6043         vrp_value_range_pool data member directly.
6044         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6045         to use extract_range_from_stmt method to avoid need for
6046         extract_range_from_assignment method.
6047         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6048         method rather than setting values_propgated data member directly.
6049         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6050         and values propagated data members and extract_range_from_assignment
6051         method.  Reorder private data members to conform to standards.
6052         Add new methods set_lattice_propagation_complete and
6053         allocate_value_range.
6055 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6057         PR rtl-optimization/83030
6058         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6059         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6060         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6061         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6062         CROSSING_JUMP_P flag is already set before setting it.
6063         * cfgrtl.c (fixup_partition_crossing): Likewise.
6064         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6065         insn as useless.
6067 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6069         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6070         Handle the case where both arguments are using gen_const_vec_series.
6072 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6074         PR c++/62170
6075         * pretty-print.c (pp_format): Move quoting implementation to
6076         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6077         to pass address of "quote" local.
6078         (pp_begin_quote): New function.
6079         (pp_end_quote): New function.
6080         * pretty-print.h (printer_fn): Convert penultimate param from bool
6081         to bool *.
6082         (pp_begin_quote): New decl.
6083         (pp_end_quote): New decl.
6084         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6085         param from bool to bool *.
6086         * tree-diagnostic.h (default_tree_printer): Likewise.
6088 2017-11-22  Jeff Law  <law@redhat.com>
6090         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6091         Initialize vr_values.
6092         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6093         vr_values via vr_values class instance rather than delegators.
6094         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6095         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6096         (evrp_range_analyzer::push_value_range): Likewise.
6097         (evrp_range_analyzer::pop_value_range): Likewise.
6098         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6099         most delegators.  Those remaining are exposed as public interfaces.
6100         Make vr_values a pointer and private.
6101         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6102         vr_values.
6103         (evrp_range_analyzer::get_vr_value): New method.
6104         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6105         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6106         (class evrp_dom_walker): Attach evrp_folder class, initialize
6107         it in the ctor.  Remove temporary delegators.
6108         (evrp_dom_walker::before_dom_children): Call methods in attached
6109         evrp_range_analyzer class via class instance pointer.  Use
6110         free value_range_constant_singleton to remove need for
6111         op_with_constant_singleton_value delegator method.  Do not
6112         create a vrp_prop class instance for every call!  Narrow
6113         scope of a couple variables.
6114         (evrp_dom_walker::cleanup): Call methods in attached
6115         evrp_range_analyzer class via class instance pointer.
6116         * vr-values.h (class vr_values): Privatize many methods and
6117         data members.
6119 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6121         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6122         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6123         * expr.c (expand_expr_real_2): Likewise.
6124         * fold-const.c (const_unop): Likewise.
6125         * optabs-tree.c (optab_for_tree_code): Likewise.
6126         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6127         * tree-inline.c (estimate_operator_cost): Likewise.
6128         * tree-pretty-print.c (dump_generic_node): Likewise.
6129         (op_code_prio): Likewise.
6130         (op_symbol_code): Likewise.
6131         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6132         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6133         * internal-fn.c (direct_internal_fn_optab): New function.
6134         (direct_internal_fn_array, direct_internal_fn_supported_p
6135         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6136         * fold-const-call.c (fold_const_reduction): New function.
6137         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6138         CFN_REDUC_MIN.
6139         * tree-vect-loop.c: Include internal-fn.h.
6140         (reduction_code_for_scalar_code): Rename to...
6141         (reduction_fn_for_scalar_code): ...this and return an internal
6142         function.
6143         (vect_model_reduction_cost): Take an internal_fn rather than
6144         a tree_code.
6145         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6146         than assignments.
6147         (vectorizable_reduction): Use internal functions rather than tree
6148         codes for the reduction operation.  Update calls to the functions
6149         above.
6150         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6151         Use calls to internal functions rather than REDUC tree codes.
6152         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6154 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6156         * config/vxworks.c (vxworks_override_options): Pick default
6157         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6158         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6159         DWARF_GNAT_ENCODINGS_DEFAULT.
6160         * config/vxworksae.h: Likewise.
6162 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6164         PR tree-optimization/83104
6165         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6166         not result.
6168 2017-11-22  Marek Polacek  <polacek@redhat.com>
6169             H.J. Lu  <hongjiu.lu@intel.com>
6170             Jason Merrill  <jason@redhat.com>
6172         PR c++/60336
6173         PR middle-end/67239
6174         PR target/68355
6175         * calls.c (initialize_argument_information): Call
6176         warn_parameter_passing_abi target hook.
6177         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6178         argument onto stack.
6179         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6180         * common.opt: Update -fabi-version description.
6181         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6182         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6183         int_size_in_bytes.
6184         (ix86_is_empty_record): New function.
6185         (ix86_warn_parameter_passing_abi): New function.
6186         (TARGET_EMPTY_RECORD_P): Redefine.
6187         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6188         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6189         * doc/tm.texi: Regenerated.
6190         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6191         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6192         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6193         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6194         instead of int_size_in_bytes.
6195         * expr.c (copy_blkmode_to_reg): Likewise.
6196         * function.c (aggregate_value_p): Return 0 for empty types.
6197         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6198         target hook.
6199         (locate_and_pad_parm): Call arg size_in_bytes instead
6200         size_in_bytes.
6201         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6202         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6203         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6204         hooks.
6205         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6206         (std_gimplify_va_arg_expr): Skip empty records.  Call
6207         arg_size_in_bytes instead size_in_bytes.
6208         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6209         * tree-core.h (tree_type_common): Add empty_flag.
6210         (tree_decl_common): Update comments.
6211         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6212         DECL_PADDING_P.
6213         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6214         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6215         DECL_PADDING_P.
6216         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6217         * tree.c (default_is_empty_type): New function.
6218         (default_is_empty_record): New function.
6219         (arg_int_size_in_bytes): New function.
6220         (arg_size_in_bytes): New function.
6221         (get_ultimate_context): New function.
6222         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6223         TRANSLATION_UNIT_WARN_EMPTY_P.
6224         (default_is_empty_record, arg_int_size_in_bytes,
6225         arg_size_in_bytes, get_ultimate_context): Declare.
6227 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6229         * config/arm/arm.c (cmse_clear_registers): New function.
6230         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6231         code by call to cmse_clear_registers.
6232         (cmse_nonsecure_entry_clear_before_return): Likewise.
6234 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6236         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6237         (vdot_s32, vdotq_s32): New.
6238         (vdot_lane_u32, vdotq_lane_u32): New.
6239         (vdot_lane_s32, vdotq_lane_s32): New.
6242 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6244         PR middle-end/82547
6245         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6246         for unsigned values with fewer HWIs than the precision.
6247         (test_overflow): New function.
6248         (wide_int_cc_tests): Call it.
6250 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6252         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6253         has the same mode class as Pmode.
6255 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6257         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6258         padding_bits_to_clear_ptr.
6259         (cmse_nonsecure_entry_clear_before_return): Likewise.
6261 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6263         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6264         auto_sbitap instead of integer bitfield to control register needing
6265         clearing.
6267 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6269         PR tree-optimization/83044
6270         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6271         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6272         that need the upper bound.  Subtract offset from
6273         get_addr_base_and_unit_offset only if positive and subtract it
6274         before division by eltsize rather than after it.
6276         PR debug/83084
6277         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6278         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6279         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6280         asm.
6282         PR middle-end/82875
6283         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6284         expand_binop with *mul_widen_optab, make sure at least one of the
6285         operands doesn't have VOIDmode.
6287         PR debug/83034
6288         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6290         PR rtl-optimization/82044
6291         PR tree-optimization/82042
6292         * dse.c (record_store): Check for overflow.
6293         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6294         add_wild_read instead of clear_rhs_from_active_local_stores on
6295         overflow and log it into dump_file.
6297 2017-11-22  Richard Biener  <rguenther@suse.de>
6299         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6300         fake edges to exit when looking for a place to insert.
6301         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6302         and remove.
6303         (insert_into_preds_of_block): Commit edge insertion immediately,
6304         assert that doesn't require new BBs.
6305         (fini_pre): Release expressions.
6306         (pass_pre::execute): Shuffle things around a bit, if the fn
6307         is too large do not compute AVAIL either as this is really the
6308         quadratic bit.
6310 2017-11-22  Richard Biener  <rguenther@suse.de>
6312         PR tree-optimization/83089
6313         * tree-if-conv.c (pass_if_conversion::execute): If anything
6314         changed reset SCEV and free the number of iteration estimates.
6316 2017-11-21  Martin Sebor  <msebor@redhat.com>
6318         PR tree-optimization/82945
6319         * calls.h (warn_nonstring_bound): Remove unused function.
6321 2017-11-21  Martin Sebor  <msebor@redhat.com>
6323         PR tree-optimization/82945
6324         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6325         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6326         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6327         functions.
6328         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6329         * calls.h (get_attr_nonstring_decl): Declare new function.
6330         * doc/extend.texi (attribute nonstring): Update.
6331         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6332         get_attr_nonstring_decl and handle it.
6333         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6334         detection of nul-termination.
6335         (strlen_to_stridx): Change to a pointer.
6336         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6337         (pass_strlen::execute): Same.
6339 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6341         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6342         for the new option -mprefer-vector-width=[none|128|256|512].
6343         * config/i386/i386.c (ix86_target_string): remove old style options
6344         -mprefer-avx256 and make -mprefer-avx128 as alias.
6345         (ix86_option_override_internal):  Apply defaults for the
6346         -mprefer-vector-width=[128|256] option.
6347         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6348         Implement macros to work with -mprefer-vector-width=.
6349         * config/i386/i386.opt: Implemented option
6350         -mprefer-vector-width=[none|128|256|512].
6351         * doc/invoke.texi: Documentation for
6352         -mprefer-vector-width=[none|128|256|512].
6354 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6356         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6358 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6360         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6362 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6364         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6365         POINTER_PLUS_EXPR.
6366         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6367         * expr.c (expand_expr_real_2): Likewise.
6368         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6369         fold_binary_loc): Likewise.
6370         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6371         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6372         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6373         MINUS_EXPR transformations.
6374         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6375         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6376         * tree-inline.c (estimate_operator_cost): Likewise.
6377         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6378         op_symbol_code): Likewise.
6379         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6380         * vr-values.c (extract_range_from_binary_expr): Likewise.
6381         * varasm.c (initializer_constant_valid_p_1): Likewise.
6382         * tree.def: New tree code POINTER_DIFF_EXPR.
6384 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6386         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6387         integer suffix to movbe mnemonic.
6388         (*bswaphi2_movbe): Ditto.
6389         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6391 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6393         PR c++/83045
6394         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6395         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6396         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6398 2017-11-21  Martin Liska  <mliska@suse.cz>
6400         * tree-inline.c (expand_call_inline): Remove not needed
6401         xstrdup_for_dump.
6403 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6404             Jakub Jelinek  <jakub@redhat.com>
6406         PR target/82880
6407         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6408         Remove static keyword from f variable.
6410 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6412         PR tree-optimization/83086
6413         * gimple-ssa-store-merging.c
6414         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6415         rather than n.base_addr.
6417 2017-11-21  Martin Liska  <mliska@suse.cz>
6419         PR rtl-optimization/82044
6420         PR tree-optimization/82042
6421         * dse.c (check_mem_read_rtx): Check for overflow.
6423 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6425         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6426         typo in comment.
6428 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6430         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6431         and its corresponding call arg location note.
6433 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6434             Andrew Burgess  <andrew.burgess@embecosm.com>
6436         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6437         declaration.
6438         (arc_return_slot_offset): Likewise.
6439         (arc_eh_return_address_location): New declaration.
6440         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6441         (MUST_SAVE_REGISTER): Add exception handler case.
6442         (MUST_SAVE_RETURN_ADDR): Likewise.
6443         (arc_frame_pointer_required): Likewise.
6444         (arc_frame_pointer_needed): New function.
6445         (arc_compute_frame_size): Changed.
6446         (arc_expand_prologue): Likewise.
6447         (arc_expand_epilogue): Likewise.
6448         (arc_initial_elimination_offset): Likewise.
6449         (arc_return_slot_offset): Delete.
6450         (arc_eh_return_address_location): New function.
6451         (arc_builtin_setjmp_frame_value): Likewise.
6452         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6453         (EH_RETURN_STACKADJ_RTX): Define.
6454         (EH_RETURN_HANDLER_RTX): Likewise.
6455         * config/arc/arc.md (eh_return): Delete.
6457 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6459         * print-tree.h (debug_vec_tree): Remove prototype.
6460         * gdbinit.in (pvt): Remove macro.
6462 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6464         PR tree-optimization/83047
6465         * gimple-ssa-store-merging.c
6466         (imm_store_chain_info::output_merged_store): If the loads with the
6467         same vuse are in different basic blocks, for load_gsi pick a load
6468         location that is dominated by the other loads.
6470         PR c++/83059
6471         * config/i386/i386.c (ix86_memmodel_check): Start
6472         -Winvalid-memory-model diagnostics with lowercase letter.
6474         PR debug/82718
6475         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6476         set in_cold_section_p to the partition containing loc_list->first.
6477         When seeing loc_list->last_before_switch node, update secname and
6478         perform range_across_switch second partition handling only after that.
6480         PR debug/82933
6481         * run-rtl-passes.c: Include debug.h.
6482         (run_rtl_passes): Call debug_hooks->assembly_start.
6483         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6484         multiple times.
6486         PR target/82981
6487         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6488         OPTAB_DIRECT in calls to expand_simple_binop.
6490 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6492         PR c/81404
6493         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6495 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6497         PR c++/72786
6498         * spellcheck.h (best_match::blithely_get_best_candidate): New
6499         accessor.
6501 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6503         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6504         with lower case letter.
6506 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6508         * config/i386/i386.md (bswaphi2): New expander.
6509         (*bswaphi2_movbe): New insn pattern.
6510         (bswaphi -> rorhi pepehole2): New peephole pattern.
6512 2017-11-20  Jeff Law  <law@redhat.com>
6514         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6515         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6516         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6517         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6518         gimple-ssa-evrp-analyze.h.
6520         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6521         set BB_VISITED here.
6522         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6524 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6526         * tree-predcom.c: Add general comment on Store-Store chains.
6527         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6528         flag in component.
6529         (get_chain_last_ref_at): Rename into...
6530         (get_chain_last_write_at): ...this.
6531         (get_chain_last_write_before_load): New function.
6532         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6533         CT_STORE_STORE when write reference is added.
6534         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6535         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6536         (initialize_root_vars_store_elim_1): Ditto.
6537         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6538         definition is created.
6539         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6540         chain by replacing it with dominant stored value.
6542 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6544         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6546 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6548         * vr-values.c (extract_range_from_binary_expr): Use a full range
6549         for VR_VARYING.
6551 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6553         * config/arm/arm.md (R4_REGNUM): Define constant.
6554         (nonsecure_call_internal): Remove r4 clobber.
6555         (nonsecure_call_value_internal): Likewise.
6556         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6557         clobber and resequence match_operands.
6558         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6559         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6560         (nonsecure_call_value_reg_thumb2): Likewise.
6562 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6564         PR tree-optimization/78821
6565         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6566         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6567         to the address of the base rather than the base itself.
6568         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6569         (find_bswap_or_nop_finalize): New function.
6570         (find_bswap_or_nop): Use it.
6571         (bswap_replace): Return a tree rather than bool, change first
6572         argument from gimple * to gimple_stmt_iterator, allow inserting
6573         into an empty sequence, allow ins_stmt to be NULL - then emit
6574         all stmts into gsi.  Fix up MEM_REF address gimplification.
6575         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6576         (struct store_immediate_info): Add N and INS_STMT non-static
6577         data members.
6578         (store_immediate_info::store_immediate_info): Initialize them
6579         from newly added ctor args.
6580         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6581         bitpos at the end.
6582         (stmts_may_clobber_ref_p): For stores call also
6583         refs_anti_dependent_p.
6584         (gather_bswap_load_refs): New function.
6585         (imm_store_chain_info::try_coalesce_bswap): New method.
6586         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6587         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6588         (imm_store_chain_info::output_merged_store): Fail if number of
6589         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6590         and NOP_EXPR rhs_code.
6591         (pass_store_merging::process_store): Compute n and ins_stmt, if
6592         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6593         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6594         ctor.
6595         (pass_store_merging::execute): Calculate dominators.
6597         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6598         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6599         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6600         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6601         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6602         class pass_optimize_bswap, bswap_replace,
6603         pass_optimize_bswap::execute): Moved to ...
6604         * gimple-ssa-store-merging.c: ... this file.
6605         Include optabs-tree.h.
6606         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6607         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6608         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6609         anonymous namespace, remove static keywords.
6610         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6611         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6613 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6615         PR bootstrap/83062
6616         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6618 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6620         * vec.h (debug_helper): New function.
6621         (DEFINE_DEBUG_VEC): New macro.
6622         * hash-set.h (debug_helper): New function.
6623         (DEFINE_DEBUG_HASH_SET): New macro.
6624         * cfg.c (debug_slim (edge)): New function.
6625         Call DEFINE_DEBUG_VEC for edges.
6626         Call DEFINE_DEBUG_HASH_SET for edges.
6627         * cfghooks.c (debug_slim (basic_block)): New function.
6628         Call DEFINE_DEBUG_VEC for basic blocks.
6629         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6630         * print-tree.c (debug_slim): New function to handle trees.
6631         Call DEFINE_DEBUG_VEC for trees.
6632         Call DEFINE_DEBUG_HASH_SET for trees.
6633         (debug (vec<tree, va_gc>) &): Remove.
6634         (debug (<vec<tree, va_gc>) *): Remove.
6635         * print-rtl.c (debug_slim): New function to handle const_rtx.
6636         Call DEFINE_DEBUG_VEC for rtx_def.
6637         Call DEFINE_DEBUG_VEC for rtx_insn.
6638         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6639         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6640         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6641         (debug (vec<rtx_insn *> *ptr): Remove.
6642         (debug_insn_vector): Remove.
6643         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6645 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6647         PR rtl-optimization/82020
6648         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6649         IF_THEN_ELSE condition.
6651 2017-11-19  Jeff Law  <law@redhat.com>
6653         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6654         of degenerates resulting from ignoring an edge.
6656 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6658         PR ipa/81360
6659         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6661 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6663         PR ipa/83001
6664         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6665         for uninitialied counts.
6667 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6669         PR ipa/60243
6670         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6672 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6674         PR target/82713
6675         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6677 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6679         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6680         "do while (0)".
6682 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6684         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6685         "do {} while (0)".
6687 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6689         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6690         macro body.
6692 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6694         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6695         "do {} while (0)".
6696         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6698 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6700         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6701         "do {} while (0)".
6702         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6703         semicolon after MCORE_EXPORT_NAME call.
6705 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6707         PR target/82961
6708         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6709         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6711 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6713         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6714         (cgraph_edge::clone): Cleanup updating of profile.
6715         * ipa-cp.c (update_profiling_info): Likewise.
6716         * ipa-inline-transform.c (inline_transform): Likewise.
6717         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6718         * ipa-split.c (execute_split_functions): Do not split when function
6719         is cold.
6720         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6721         * profile-count.c (profile_count::dump): Add global0.
6722         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6723         undefined.
6724         (profile_count::to_sreal_scale): Likewise.
6725         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6726         (profile_count::combine_with_ipa_count): New function.
6727         * profile-count.h (profile_guessed_global0adjusted): New.
6728         (profile_count::adjusted_zero): New.
6729         (profile_count::global0adjusted): New.
6730         (profile_count::combine_with_ipa_count): New.
6731         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6732         correct profile of return block of split functions.
6733         (copy_cfg_body): Remove unused profile_count.
6734         (copy_body): Likewise.
6735         (expand_call_inline): Update.
6736         (tree_function_versioning): Update.
6738 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6740         * hash-set.h (hash_set::empty): New.
6741         * tree-ssa-threadbackward.h: Delete.
6742         * tree-ssa-threadbackward.c (class thread_jumps): New.
6743         Move max_threaded_paths into class.
6744         (fsm_find_thread_path): Remove arguments that are now in class.
6745         (profitable_jump_thread_path): Rename to...
6746         (thread_jumps::profitable_jump_thread_path): ...this.
6747         (convert_and_register_jump_thread_path): Rename to...
6748         (thread_jumps::convert_and_register_current_path): ...this.
6749         (check_subpath_and_update_thread_path): Rename to...
6750         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6751         (register_jump_thread_path_if_profitable): Rename to...
6752         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6753         (handle_phi): Rename to...
6754         (thread_jumps::handle_phi): ...this.
6755         (handle_assignment): Rename to...
6756         (thread_jumps::handle_assignment): ...this.
6757         (fsm_find_control_statement_thread_paths): Rename to...
6758         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6759         (find_jump_threads_backwards): Rename to...
6760         (thread_jumps::find_jump_threads_backwards): ...this.
6761         Initialize path local data.
6762         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6763         from within thread_jumps class.
6764         (pass_early_thread_jumps::execute): Same.
6766 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6768         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6770 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6772         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6773         to canonical location.
6775 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6777         PR target/81356
6778         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6779         Remove.
6780         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6782 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6784         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6785         rcrt1.o%s/grcrt1.o%s for -static-pie.
6787 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6789         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6790         ix86_shift_rotate_cost): Break out from ...
6791         (ix86_rtx_costs): ... here.
6792         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6793         vector operations.
6795 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6797         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6798         when entry block was promoted unlikely.
6799         (estimate_bb_frequencies): Increase frequency scale.
6800         * profile-count.h (profile_count): Export precision info.
6802 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6804         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6805         disturbin profile of entry block.
6807 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6809         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6810         roundoff errors.
6812 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6814         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6815         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6817 2017-11-17  Jeff Law  <law@redhat.com>
6819         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6820         from evrp_dom_walker class.  Various methods moved into new class.
6821         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6822         (evrp_range_analyzer::enter): New method.
6823         (evrp_range_analyzer::leave): New method.
6824         (evrp_dom_walker): Remove delegators no longer needed by this class.
6825         Replace vr_values data member with evrp_range_analyzer
6827         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6828         method extracted from evrp_dom_walker::before_dom_children.
6829         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6830         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6832         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6833         Add private copy constructor and move assignment operators.
6834         Privatize methods and class data where trivially possible.
6835         (evrp_dom_walker::cleanup): New function, extracted from
6836         execute_early_vrp.  Simplify access to class data.
6838         * vr-values.h (get_output_for_vrp): Prototype.
6839         * vr-values.c (get_output_for_vrp): New function extracted from
6840         vrp_visit_assignment_or_call and extract_range_from_stmt.
6841         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6843 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6845         * config/aarch64/aarch64.c
6846         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6847         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6848         tune_params::AUTOPREFETCHER_WEAK.
6850 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6852         PR target/82641
6853         * config/arm/arm.c (arm_valid_target_attribute_rec):
6854         Parse "arch=" and "+<ext>".
6855         (arm_valid_target_attribute_tree): Re-init global options.
6856         (arm_option_override): Make non-static.
6857         (arm_options_perform_arch_sanity_checks): Make errors fatal.
6858         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6859         (__ARM_FEATURE_CRC32): Support undef.
6860         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6861         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
6863 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
6865         * gdbinit.in (break-on-diagnostic): New command.
6867 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6869         * config/i386/i386.c (ix86_expand_epilogue): Change simple
6870         return to indirect jump for EH return if control-flow
6871         protection is enabled. Change explicit 'false' argument in
6872         pro_epilogue_adjust_stack with a value of
6873         flag_cf_protection.
6874         * config/i386/i386.md (simple_return_indirect_internal):
6875         Remove SImode restriction to support 64-bit.
6877 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6879         * combine.c (added_notes_insn): New.
6880         (try_combine): Handle added_notes_insn like added_links_insn.
6881         Rewrite return value code.
6882         (distribute_notes): Set added_notes_insn to the earliest insn we added
6883         a note to.
6885 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6887         PR rtl-optimization/82621
6888         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
6889         dest of one of those SETs is unused.
6891 2017-11-17  Richard Biener  <rguenther@suse.de>
6893         PR fortran/83017
6894         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
6895         * tree-pretty-print.c (dump_generic_node): Handle
6896         annot_expr_parallel_kind.
6897         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
6898         * gimplify.c (gimple_boolify): Likewise.
6900 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6902         * config.gcc (extra_headers): Add cet.h for x86 targets.
6903         * config/i386/cet.h: New file.
6904         * doc/install.texi: Add --enable-cet/--disable-cet.
6906 2017-11-17  Richard Biener  <rguenther@suse.de>
6908         PR tree-optimization/83017
6909         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
6910         (gen_parallel_loop): Properly count iterations.
6911         (parallelize_loops): Handle loop->can_be_parallel independent
6912         of flag_loop_parallelize_all.  Make static profitability test match
6913         the runtime one.
6914         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
6915         * invoke.texi (parloops-min-per-thread): Document.
6917 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
6919         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
6920         upstreaming review comments.
6922 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6924         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
6925         with fast unaligned access.
6926         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
6928 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6930         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
6931         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
6933 2017-11-17  Richard Biener  <rguenther@suse.de>
6935         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
6936         folding of references.
6938 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
6940         PR middle-end/78809
6941         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
6942         of replacing call to strncmp with corresponding call to strcmp when
6943         meeting conditions.
6945 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6947         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
6948         option prefer-avx256 for skylake-avx512 configuration.
6949         * config/i386/i386.c (ix86_option_override_internal): Ditto.
6950         (get_builtin_code_for_version): Ditto.
6952 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6953             Monk Chiang  <sh.chiang04@gmail.com>
6955         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
6956         (FIXED_REGISTERS): Reserve more register numbers.
6957         (CALL_USED_REGISTERS): Likewise.
6958         (REG_ALLOC_ORDER): Likewise.
6959         (REG_CLASS_CONTENTS): Likewise.
6960         (REGISTER_NAMES): Likewise.
6962 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6963             Kito Cheng  <kito.cheng@gmail.com>
6965         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
6966         V2SI.
6967         * config/nds32/iterators.md: Add vector mode iterators and attributes.
6969 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
6971         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
6972         parameter list for vec_splats.
6974 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
6976         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
6977         date of C17.
6978         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
6980 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
6982         PR bootstrap/82856
6983         * doc/install.texi: Document incompatibility of Perl >=5.6.26
6984         with the required version of automake 1.11.6.
6986 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
6988         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
6989         for it.
6990         (DU_C2_3_power9): Correct reservation combinations.
6991         (FP_DIV_power9, VEC_DIV_power9): New.
6992         (power9-alu): Split out rotate/shift...
6993         (power9-rot): ...to here, correct dispatch resource.
6994         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
6995         resource.
6996         (power9-fp): Correct latency.
6997         (power9-sdiv): Add div/sqrt resource.
6998         (power9-ddiv): Correct latency, add div/sqrt resource.
6999         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7000         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7001         resource.
7002         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7004 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7006         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7007         switch statement mapping KF built-ins to TF built-ins if we don't
7008         have the proper ISA 3.0 assembler support.
7010 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7012         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7013         (lower_emutls_function_body): Do not compute it.
7015 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7017         * ipa-split.c (split_bb_info): Turn time to sreal.
7018         (split_point): Likewise.
7019         (dump_split_point): Likewise.
7020         (fine_split_points): Likewise.
7021         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7023 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7025         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7026         in sreal.
7028 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7030         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7031         edges.
7032         (expensive_function_p): Remove useless assert.
7033         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7035 2017-11-16  Martin Sebor  <msebor@redhat.com>
7037         PR tree-optimization/82588
7038         PR tree-optimization/82583
7039         * tree-vrp.c (check_array_ref): Handle flexible array members,
7040         string literals, and inner indices.
7041         (search_for_addr_array): Add detail to diagnostics.
7043 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7045         PR c++/82836
7046         PR c++/82737
7047         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7048         (SET_DECL_ASSEMBLER_NAME): Forward to
7049         overwrite_decl_assembler_name.
7050         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7051         (overwrite_decl_assembler_name): Declare.
7052         * tree.c (overwrite_decl_assembler_name): New.
7053         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7054         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7055         (LANG_HOOKS_INITIALIZER): Add it.
7056         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7057         * langhooks.c (lhd_set_decl_assembler_name): Use
7058         SET_DECL_ASSEMBLER_NAME.
7059         (lhd_overwrite_decl_assembler_name): Default implementation.
7061 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7062             Jackson Woodruff  <jackson.woodruff@arm.com>
7064         PR tree-optimization/71026
7065         * match.pd: Canonicalize constant multiplies in division.
7067 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7069         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7070         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7071         Remove OPT_fomit_frame_pointer entry.
7072         * common/config/arc/arc-common.c: Likewise.
7073         * common/config/arm/arm-common.c: Likewise.
7074         * common/config/avr/avr-common.c: Likewise.
7075         * common/config/c6x/c6x-common.c: Likewise.
7076         * common/config/cr16/cr16-common.c: Likewise.
7077         * common/config/cris/cris-common.c: Likewise.
7078         * common/config/epiphany/epiphany-common.c: Likewise.
7079         * common/config/fr30/fr30-common.c: Likewise.
7080         * common/config/frv/frv-common.c: Likewise.
7081         * common/config/ia64/ia64-common.c: Likewise.
7082         * common/config/iq2000/iq2000-common.c: Likewise.
7083         * common/config/lm32/lm32-common.c: Likewise.
7084         * common/config/m32r/m32r-common.c: Likewise.
7085         * common/config/mcore/mcore-common.c: Likewise.
7086         * common/config/microblaze/microblaze-common.c: Likewise.
7087         * common/config/mips/mips-common.c: Likewise.
7088         * common/config/mmix/mmix-common.c: Likewise.
7089         * common/config/mn10300/mn10300-common.c: Likewise.
7090         * common/config/nios2/nios2-common.c: Likewise.
7091         * common/config/pa/pa-common.c: Likewise.
7092         * common/config/pdp11/pdp11-common.c: Likewise.
7093         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7094         * common/config/riscv/riscv-common.c: Likewise.
7095         * common/config/rs6000/rs6000-common.c: Likewise.
7096         * common/config/rx/rx-common.c: Likewise.
7097         * common/config/s390/s390-common.c: Likewise.
7098         * common/config/sh/sh-common.c: Likewise.
7099         * common/config/sparc/sparc-common.c: Likewise.
7100         * common/config/tilegx/tilegx-common.c: Likewise.
7101         * common/config/tilepro/tilepro-common.c: Likewise.
7102         * common/config/v850/v850-common.c: Likewise.
7103         * common/config/visium/visium-common.c: Likewise.
7104         * common/config/xstormy16/xstormy16-common.c: Likewise.
7105         * common/config/xtensa/xtensa-common.c: Likewise.
7106         * invoke.texi (-fomit-frame-pointer): Update documentation.
7108 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7110         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7112 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7114         * cfg.c (scale_bbs_frequencies_int,
7115         cale_bbs_frequencies_gcov_type): Remove.
7116         * cfg.h (scale_bbs_frequencies_int,
7117         cale_bbs_frequencies_gcov_type): Remove.
7119 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7121         * tree-ssa-loop-manip.c
7122         (scale_dominated_blocks_in_loop): Update to profile counts.
7123         (tree_transform_and_unroll_loop): Likewise.
7125 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7127         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7128         scale_bbs_frequencies_int.
7130 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7132         * final.c (compute_alignments): Use counts rather than frequencies.
7134 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7136         * cfgloopanal.c: Include sreal.h
7137         (average_num_loop_insns): Use counts and sreal for accounting.
7139 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7141         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7142         manipulation.
7144 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7145             Kito Cheng  <kito.cheng@gmail.com>
7147         * config/nds32/constraints.md: Provide more constraints.
7148         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7149         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7150         support constraints usage.
7152 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7154         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7156 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7157             Kito Cheng  <kito.cheng@gmail.com>
7159         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7160         * config/nds32/nds32.opt: Refine the layout.
7161         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7162         TARGET_EXT_STRING): Support new options.
7163         * config/nds32/nds32.h: Likewise.
7164         * config/nds32/nds32.md: Likewise.
7165         * config/nds32/nds32-predicates.c: Likewise.
7166         * config/nds32/constraints.md: Likewise.
7167         * common/config/nds32/nds32-common.c: Likewise.
7169 2017-11-16  Julia Koval  <julia.koval@intel.com>
7171         PR target/82983
7172         * config/i386/gfniintrin.h: Add sse check.
7173         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7175 2017-11-16  Julia Koval  <julia.koval@intel.com>
7177         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7178         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7179         (ix86_handle_option): Handle -mavx512vbmi2.
7180         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7181         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7182         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7183         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7184         (ix86_valid_target_attribute_inner_p): Ditto.
7185         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7186         * config/i386/i386.opt (mavx512vbmi2): New option.
7187         * doc/invoke.texi: Add new option.
7189 2017-11-16  Julia Koval  <julia.koval@intel.com>
7191         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7192         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7193         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7194         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7195         _mm512_gf2p8mul_epi8): New intrinsics.
7196         * config/i386/i386-builtin-types.def
7197         (V64QI_FTYPE_V64QI_V64QI): New type.
7198         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7199         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7200         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7201         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7202         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7203         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7205 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7207         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7208         explicitly as a stream of bytes.
7210 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7212         * config/rs6000/altivec.h (vec_xst_be): New #define.
7213         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7214         and externalize from *altivec_vperm_<mode>_internal.
7215         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7216         instantiation.
7217         (XL_BE_V8HI): Likewise.
7218         (XL_BE_V4SI): Likewise.
7219         (XL_BE_V4SI): Likewise.
7220         (XL_BE_V2DI): Likewise.
7221         (XL_BE_V4SF): Likewise.
7222         (XL_BE_V2DF): Likewise.
7223         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7224         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7225         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7226         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7227         VSX_BUILTIN_VEC_XST_BE.
7228         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7229         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7230         built-ins.
7231         (altivec_init_builtins): Replace conditional calls to def_builtin
7232         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7233         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7234         with unconditional calls.  Remove calls to def_builtin for
7235         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7236         __builtin_vec_xst_be.
7237         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7238         to define_expand, and add alternate RTL generation for P8.
7239         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7240         vsx_ld_elemrev_v8hi.
7241         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7242         add alternate RTL generation for P8.
7243         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7244         vsx_ld_elemrev_v16qi.
7245         (vsx_st_elemrev_v8hi): Convert define_insn
7246         to define_expand, and add alternate RTL generation for P8.
7247         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7248         vsx_st_elemrev_v8hi.
7249         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7250         add alternate RTL generation for P8.
7251         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7252         vsx_st_elemrev_v16qi.
7254 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7256         PR target/82990
7257         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7258         TARGET_AVX512ER check.
7259         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7260         neither -mzeroupper nor -mno-zeroupper is used and
7261         TARGET_EMIT_VZEROUPPER is set.
7262         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7263         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7265 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7267         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7268         folding of vector compares.
7269         (fold_build_vec_cmp): New helper function.
7270         (fold_compare_helper): New helper function.
7271         (builtin_function_type): Add compare builtins to the list of functions
7272         having unsigned arguments.  Cosmetic updates to comment indentation.
7273         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7274         the not+eq combination.
7276 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7278         PR tree-optimization/82726
7279         PR tree-optimization/70754
7280         * tree-predcom.c (order_drefs_by_pos): New function.
7281         (combine_chains): Move code setting has_max_use_after to...
7282         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7283         according to position information.
7284         (tree_predictive_commoning_loop): Update call to above function.
7285         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7287 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7289         PR tree-optimization/82726
7290         Revert
7291         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7293         PR tree-optimization/70754
7294         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7295         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7296         combined stmt before it if not NULL.
7297         (combine_chains): Process refs reversely and compute dominance point
7298         for root ref.
7300         Revert
7301         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7303         PR tree-optimization/79663
7304         * tree-predcom.c (combine_chains): Process refs in reverse order
7305         only for ZERO length chains, and add explaining comment.
7307 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7309         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7310         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7311         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7313 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7315         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7317 2017-11-15  Martin Liska  <mliska@suse.cz>
7319         * tree-cfg.c (pass_warn_function_return::execute):
7320         Compare warn_return_type for greater than zero.
7322 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7324         PR target/82941
7325         PR target/82942
7326         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7327         to return true on Xeon and not on Xeon Phi.
7328         (ix86_check_avx256_register): Changed to ...
7329         (ix86_check_avx_upper_register): ... this. Add extra check for
7330         VALID_AVX512F_REG_OR_XI_MODE.
7331         (ix86_avx_u128_mode_needed): Changed
7332         ix86_check_avx256_register to ix86_check_avx_upper_register.
7333         (ix86_check_avx256_stores): Changed to ...
7334         (ix86_check_avx_upper_stores): ... this. Changed
7335         ix86_check_avx256_register to ix86_check_avx_upper_register.
7336         (ix86_avx_u128_mode_after): Changed
7337         avx_reg256_found to avx_upper_reg_found. Changed
7338         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7339         (ix86_avx_u128_mode_entry): Changed
7340         ix86_check_avx256_register to ix86_check_avx_upper_register.
7341         (ix86_avx_u128_mode_exit): Ditto.
7342         * config/i386/i386.h: (host_detect_local_cpu): New define.
7344 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7346         * config/arm/xgene1.md (xgene1): Split into automatons
7347         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7348         (xgene1_f_load): Adjust reservations and/or types.
7349         (xgene1_f_store): Likewise.
7350         (xgene1_load_pair): Likewise.
7351         (xgene1_store_pair): Likewise.
7352         (xgene1_fp_load1): Likewise.
7353         (xgene1_load1): Likewise.
7354         (xgene1_store1): Likewise.
7355         (xgene1_move): Likewise.
7356         (xgene1_alu): Likewise.
7357         (xgene1_simd): Likewise.
7358         (xgene1_bfm): Likewise.
7359         (xgene1_neon_load1): Likewise.
7360         (xgene1_neon_store1): Likewise.
7361         (xgene1_neon_logic): Likewise.
7362         (xgene1_neon_st1): Likewise.
7363         (xgene1_neon_ld1r): Likewise.
7364         (xgene1_alu_cond): Added.
7365         (xgene1_shift_reg): Likwise.
7366         (xgene1_bfx): Likewise.
7367         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7369 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7371         PR target/82981
7372         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7373         ssa-iterators.h.
7374         (can_widen_mult_without_libcall): New function.
7375         (expand_mul_overflow): If only checking unsigned mul overflow,
7376         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7377         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7378         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7379         (expand_DIVMOD): Formatting fix.
7380         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7381         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7382         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7384         PR tree-optimization/82977
7385         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7386         constructed temporary to strlen_to_stridx.put.
7388 2017-11-15  Martin Liska  <mliska@suse.cz>
7390         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7391         * configure: Regenerate.
7393 2017-11-15  Martin Liska  <mliska@suse.cz>
7395         PR target/82927
7396         * config/sh/sh-mem.cc: Use proper probability for
7397         REG_BR_PROB_NOTE.
7399 2017-11-14  Jeff Law  <law@redhat.com>
7401         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7402         the red zone for stack_clash_protection_final_dynamic_probe targets
7403         when the total dynamic stack size is zero bytes.
7405         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7406         blocks is post order.
7408 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7410         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7411         * final.c (rest_of_clean_state): Set it for the
7412         -fcompare-debug dump.
7413         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7414         class when TDF_COMPARE_DEBUG is set.
7416         * dwarf2out.c (gen_producer_string): Discard
7417         OPT_fcompare_debug.
7419 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7421         PR c/81156
7422         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7423         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7424         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7425         (__TGMATH_REAL_2_3): Remove macros.
7426         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7427         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7428         __builtin_tgmath.
7429         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7430         __TGMATH_REAL_2.
7431         (remquo): Define using __TGMATH_REAL_3.
7433 2017-11-14  Jeff Law  <law@redhat.com>
7435         * vr-values.c: New file with contents extracted from tree-vrp.c.
7436         * Makefile.in (OBJS): Add vr-values.o
7437         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7438         (set_value_range, set_and_canonicalize_value_range): Likewise.
7439         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7440         (value_range_constant_singleton, symbolic_range_p): Likewise.
7441         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7442         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7443         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7444         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7445         (find_case_label_range, find_case_label_index): Likewise.
7446         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7447         (range_int_cst_singleton_p, value_inside_range): Likewise.
7448         (get_single_symbol): Likewise.
7449         (switch_update): Move structure definition here.
7450         (to_remove_edges, to_update_switch_stmts): Provide externs.
7451         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7452         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7453         (vrp_val_is_min, set_value_range): Likewise.
7454         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7455         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7456         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7457         (range_is_nonnull, range_int_cst_p): Likewwise.
7458         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7459         (get_single_symbol, operand_less_p): Likewise
7460         (compare_values_warnv, compare_values): Likewise.
7461         (value_inside_range, value_range_constant_singleton): Likewise.
7462         (zero_nonzero_bitgs_from_vr): Likewise.
7463         (extract_range_from_binary_expr_1): Likewise.
7464         (overflow_comparison_p): Likewise.
7465         (to_remove_edges, to_update_switch_stmts): Likewise.
7466         (find_case_label-index, find_case_label_range): Likewise.
7467         (switch_update, set_value_range_to_nonnegative): Remove.
7468         (set_value_range_to_truthvalue): Likewise.
7469         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7470         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7471         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7472         (find_case_label_ranges, test_for_singularity): Likewise.
7473         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7474         (x_vr_values): Move to its remaining use site.
7476 2017-11-10  Jeff Law  <law@redhat.com>
7478         * vr-values.h (VR_INITIALIZER): Move #define here.
7479         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7480         * Makefile.in (OBJS): Add tree-evrp.o
7481         * tree-vrp.h (assert_info): Move structure definition here.
7482         (set_value_range_to_varying): Prototype.
7483         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7484         (infer_value_range, register_edge_assert_for): Likewise.
7485         (stmt_interesting_for_vrp): Likewise.
7486         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7487         (set_value_range_to_varying): No longer static.
7488         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7489         (infer_value_range, register_edge_assert_for): Likewise.
7491 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7493         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7494         generate the XXBRD instruction.
7496         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7497         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7498         as being compatible if -mabi=ieeelongdouble.
7499         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7500         to setup float128 built-ins with hardware support.
7501         (BU_FLOAT128_HW_2): Likewise.
7502         (BU_FLOAT128_HW_3): Likewise.
7503         (BU_FLOAT128_HW_VSX_1): Likewise.
7504         (BU_FLOAT128_HW_VSX_2): Likewise.
7505         (scalar_extract_expq): Change float128 built-in functions to
7506         accommodate having both KFmode and TFmode functions.  Use the
7507         KFmode variant as the default.
7508         (scalar_extract_sigq): Likewise.
7509         (scalar_test_neg_qp): Likewise.
7510         (scalar_insert_exp_q): Likewise.
7511         (scalar_insert_exp_qp): Likewise.
7512         (scalar_test_data_class_qp): Likewise.
7513         (sqrtf128_round_to_odd): Delete processing the round to odd
7514         built-in functions as special built-in functions, and define them
7515         as float128 built-ins.  Use the KFmode variant as the default.
7516         (truncf128_round_to_odd): Likewise.
7517         (addf128_round_to_odd): Likewise.
7518         (subf128_round_to_odd): Likewise.
7519         (mulf128_round_to_odd): Likewise.
7520         (divf128_round_to_odd): Likewise.
7521         (fmaf128_round_to_odd): Likewise.
7522         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7523         support for KFmode and TFmode xststdcqp calls.
7524         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7525         point, switch the built-in handlers for the get/set float128
7526         exponent, get float128 mantissa, float128 test built-ins, and the
7527         float128 round to odd built-in functions.  Eliminate creating the
7528         float128 round to odd built-in functions as special built-ins.
7529         (rs6000_init_builtins): Eliminate special creation of the float128
7530         round to odd built-in functions.
7531         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7532         function insns to support both TFmode and KFmode variants.
7533         (xsxsigqp_<mode>): Likewise.
7534         (xsiexpqpf_<mode>): Likewise.
7535         (xsiexpqp_<mode>): Likewise.
7536         (xststdcqp_<mode>): Likewise.
7537         (xststdcnegqp_<mode>): Likewise.
7538         (xststdcqp_<mode>): Likewise.
7540 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7542         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7543         unused path_in_freq_ptr parameter.
7544         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7546 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7548         * ipa-inline.c (edge_badness): Dump sreal frequency.
7549         (compute_inlined_call_time): Match natural implementaiton ...
7550         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7551         forgotten division by CGRAPH_FREQ_BASE.
7553 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7555         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7556         Solaris 11.  Update comment.
7557         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7558         renaming.
7559         * config/sol2.h (STARTFILE_SPEC): Likewise.
7560         * configure: Regenerate.
7562 2017-11-14  Carl Love  <cel@us.ibm.com>
7564         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7565         le_ and be_ prefixes to swap* variables.  Remove
7566         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7567         statements.
7569 2017-11-14  Jason Merrill  <jason@redhat.com>
7571         Support GTY((cache)) on hash_map.
7572         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7573         (ggc_cache_remove): Override it instead of ggc_mx.
7574         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7575         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7576         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7577         (simple_cache_map_traits): Override maybe_mx.
7578         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7579         (hash_map): Friend gt_cleare_cache.
7580         (gt_cleare_cache): New.
7581         * tree.h (tree_cache_traits): New hash_map traits class.
7582         (tree_cache_map): New typedef.
7584 2017-11-14  Richard Biener  <rguenther@suse.de>
7586         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7587         paramter and handling.
7588         (cleanup_control_flow_bb): Likewise.
7589         (cleanup_control_flow_pre): New helper performing a DFS walk
7590         to call cleanup_control_flow_bb in PRE order.
7591         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7592         via cleanup_control_flow_pre.
7594 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7596         * config/aarch64/aarch64-simd.md
7597         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7598         (*aarch64_simd_bsl<mode>_alt): Likewise.
7599         (aarch64_simd_bsldi_internal): New.
7600         (aarch64_simd_bsldi_alt): Likewise.
7602 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7604         * tracer.c (better_p): Do not compare frequencies.
7605         * reg-stack.c (better_edge): Likewise.
7606         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7607         and back.
7609 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7611         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7612         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7613         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7614         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7615         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7616         * predict.c (maybe_hot_frequency_p): Inline to ...
7617         (maybe_hot_count_p): ... here; rewrite to counts.
7618         (counts_to_freqs): Rename to ...
7619         (update_max_bb_count): ... this one.
7620         (expensive_function_p): Use counts.
7621         (estimate_bb_frequencies): Update.
7622         (rebuild_frequencies): Update.
7623         * predict.h (counts_to_freqs): Rename to ...
7624         (update_max_bb_count): ... this one.
7625         * profile.c (compute_branch_probabilities): Add debug info
7626         * tree-inline.c (expand_call_inline): Update debug info.
7627         (optimize_inline_calls): Use update_max_bb_count..
7628         (tree_function_versioning): Use update_max_bb_count..
7629         * value-prof.c (gimple_value_profile_transformations):
7630         Do not use update_max_bb_count.
7632 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7634         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7635         always use frequencies.
7637 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7639         * bb-reorder.c: Remove frequencies from comments.
7640         (better_edge_p): Use profile counts.
7641         (find_traces): Dump profile counts.
7642         (rotate_loop): Use profile counts.
7643         (find_traces_1_round): Likewise.
7644         (connect_better_edge_p): Use counts instead of probabilities for
7645         reverse walk.
7646         (copy_bb_p): Drop early check for non-0 frequency.
7647         (sanitize_hot_paths): Update comments.
7649 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7651         * ipa-split.c (struct split_point): Add count.
7652         (consider_split): Do not compute incoming frequency; compute incoming
7653         count and store it to split_point.
7654         (split_function): Set count of the call to split part correctly.
7656 2017-11-13  Carl Love  <cel@us.ibm.com>
7658         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7660 2017-11-13  Tom Tromey  <tom@tromey.com>
7662         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7664 2017-11-13  Carl Love  <cel@us.ibm.com>
7666         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7667         Add support for builtins:
7668         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7669         vector {un,}signed {char,int,short},
7670         vector {un,}signed {char,int,short}) arguments.
7671         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7672         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7673         Add BU_P9V_AV_2 expansions for the builtins.
7674         * config/rs6000/altivec.h (vec_first_match_index,
7675         vec_first_mismatch_index, vec_first_match_or_eos_index,
7676         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7677         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7678         new extern declaration.
7679         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7680         * config/rs6000/vsx.md (first_match_index_<mode>,
7681         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7682         first_mismatch_or_eos_index_<mode>): Add define expand.
7683         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7684         * doc/extend.texi: Update the built-in documenation file for the new
7685         built-in functions.
7687 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7689         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7690         into the min/max operations for _Float<N> and _Float<N>X types.
7692 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7694         PR lto/81351
7695         * dwarf2out.c (do_eh_frame): New static variable.
7696         (dwarf2out_begin_prologue): Set it.
7697         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7699 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7701         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7703         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7705 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7707         PR lto/81351
7708         * debug.h (dwarf2out_do_eh_frame): Declare.
7709         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7710         (dwarf2out_do_frame): Use it.
7711         (dwarf2out_do_cfi_asm): Likewise.
7712         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7713         (dwarf2out_assembly_start): Likewise.
7714         (dwarf2out_begin_prologue): Fix comment.
7715         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7716         if the target needs either debug or unwind DWARF2 info.
7717         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7718         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7719         -ftrapping-math, -ftrapv and -fwrapv.
7721 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7723         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7724         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7725         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7726         (estimate_edge_size_and_time): Likewise.
7727         (ipa_merge_fn_summary_after_inlining): Likewise.
7728         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7729         (compute_uninlined_call_time): Use sreal_frequency.
7730         (compute_inlined_call_time): Likewise.
7731         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7732         * profile-count.c: Include sreal.h.
7733         (profile_count::to_sreal_scale): New.
7734         * profile-count.h: Forward declare sreal.
7735         (profile_count::to_sreal_scale): Declare.
7737 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7739         * diagnostic.c (maybe_line_and_column): New.
7740         (diagnostic_get_location_text): Use it.
7741         (diagnostic_report_current_module): Likewise.
7742         (test_diagnostic_get_location_text): Add tests.
7744 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7746         * doc/md.texi (Specifying processor pipeline description): Fix
7747         incorrect latency for the div instruction example.
7749 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7751         PR tree-optimization/78821
7752         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7753         that bit_not_p is the same.
7754         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7755         (split_group): Count precisely bit_not_p bits in each statement.
7756         (invert_op): New function.
7757         (imm_store_chain_info::output_merged_store): Use invert_op to
7758         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7759         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7761 2017-11-13  Martin Liska  <mliska@suse.cz>
7763         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7764         (struct source_info): Likewise.
7765         (add_branch_counts): Likewise.
7766         (add_line_counts): Likewise.
7767         (function_summary): Likewise.
7768         (output_intermediate_line): Likewise.
7769         (generate_results): Likewise.
7771 2017-11-13  Martin Liska  <mliska@suse.cz>
7773         * gcov.c (struct block_info): Remove typedef for block_t.
7774         (struct line_info): Likewise.
7775         (line_info::has_block): Likewise.
7776         (EXIT_BLOCK): Likewise.
7777         (unblock): Likewise.
7778         (circuit): Likewise.
7779         (get_cycles_count): Likewise.
7780         (process_file): Likewise.
7781         (read_graph_file): Likewise.
7782         (solve_flow_graph): Likewise.
7783         (find_exception_blocks): Likewise.
7784         (add_line_counts): Likewise.
7785         (accumulate_line_info): Likewise.
7786         (output_line_details): Likewise.
7788 2017-11-13  Martin Liska  <mliska@suse.cz>
7790         * gcov.c (struct arc_info): Remove typedef for arc_t.
7791         (struct line_info): Likewise.
7792         (add_branch_counts): Likewise.
7793         (output_branch_count): Likewise.
7794         (function_info::~function_info): Likewise.
7795         (circuit): Likewise.
7796         (output_intermediate_line): Likewise.
7797         (read_graph_file): Likewise.
7798         (solve_flow_graph): Likewise.
7799         (find_exception_blocks): Likewise.
7800         (add_line_counts): Likewise.
7801         (accumulate_line_info): Likewise.
7802         (output_line_details): Likewise.
7803         (output_function_details): Likewise.
7805 2017-11-13  Martin Liska  <mliska@suse.cz>
7807         * gcov.c (struct function_info): Remove typedef for function_t.
7808         (struct source_info): Likewise.
7809         (source_info::get_functions_at_location): Likewise.
7810         (solve_flow_graph): Likewise.
7811         (find_exception_blocks): Likewise.
7812         (add_line_counts): Likewise.
7813         (output_intermediate_file): Likewise.
7814         (process_file): Likewise.
7815         (generate_results): Likewise.
7816         (release_structures): Likewise.
7817         (read_graph_file): Likewise.
7818         (read_count_file): Likewise.
7819         (accumulate_line_counts): Likewise.
7820         (output_lines): Likewise.
7822 2017-11-13  Martin Liska  <mliska@suse.cz>
7824         * gcov.c (function_info::function_info): Remove num_counts
7825         and add vector<gcov_type>.
7826         (function_info::~function_info): Use the vector.
7827         (process_file): Likewise.
7828         (read_graph_file): Likewise.
7829         (read_count_file): Likewise.
7830         (solve_flow_graph): Likewise.
7832 2017-11-13  Martin Liska  <mliska@suse.cz>
7834         * gcov.c (function_info::is_artificial): New function.
7835         (process_file): Erase all artificial early.
7836         (generate_results): Skip as all artificial are already
7837         removed.
7839 2017-11-13  Martin Liska  <mliska@suse.cz>
7841         * gcov.c (read_graph_file): Store to global vector of functions.
7842         (read_count_file): Iterate the vector.
7843         (process_file): Likewise.
7844         (generate_results): Likewise.
7845         (release_structures): Likewise.
7847 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7849         PR tree-optimization/82954
7850         * gimple-ssa-store-merging.c
7851         (imm_store_chain_info::coalesce_immediate_stores): If
7852         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7854 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
7856         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7857         Upddate call to ENDIAN_LANE_N.
7858         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7859         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7860         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7861         and use aarch64_endian_lane_rtx.
7862         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
7864 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7866         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
7868 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7870         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
7872 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7874         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
7875         body.
7876         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7877         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7878         * defaults.h (ASM_OUTPUT_LABELREF): Same.
7880 2017-11-11  Martin Sebor  <msebor@redhat.com>
7882         PR c/81117
7883         * doc/extend.texi (attribute nonstring): Remove spurious argument.
7885         PR bootstrap/82948
7886         * prefic.c (translate_name): Replace strncpy with memcpy to
7887         avoid -Wstringop-truncation.
7889 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7891         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
7893 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7895         * predict.c (maybe_hot_frequency_p): Do not use cfun.
7897 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7899         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
7900         merging.
7902 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
7904         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
7905         enable generating XXBRH if the value is in a vector register.
7906         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
7907         value is in a vector register.
7908         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
7909         register to register bswap64's instead of doing the GPR sequence
7910         used on previous machines.
7911         (bswapdi2_xxbrd): New insn.
7912         (bswapdi2_reg): Disallow on ISA 3.0.
7913         (register to register bswap64 splitter): Do not split the insn on
7914         ISA 3.0 systems that use XXBRD.
7916 2017-11-10  Martin Sebor  <msebor@redhat.com>
7918         PR c/81117
7919         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
7920         with memcpy.
7921         (find_subframework_file): Same.
7923 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7925         * auto-profile.c (afdo_indirect_call): Drop frequency.
7926         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
7927         (cgraph_node::create_edge): Drop frequency argument.
7928         (cgraph_node::create_indirect_edge): Drop frequency argument.
7929         (cgraph_edge::make_speculative): Drop frequency arguments.
7930         (cgraph_edge::resolve_speculation): Do not update frequencies
7931         (cgraph_edge::dump_edge_flags): Do not dump frequency.
7932         (cgraph_node::dump): Check consistency in IPA mode.
7933         (cgraph_edge::maybe_hot_p): Use IPA counter.
7934         (cgraph_edge::verify_count_and_frequency): Rename to ...
7935         (cgraph_edge::verify_count): ... this one; drop frequency checking.
7936         (cgraph_node::verify_node): Update.
7937         * cgraph.h (struct cgraph_edge): Drop frequency.
7938         (cgraph_edge::frequency): New function.
7939         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
7940         frequencies.
7941         (cgraph_edge::rebuild_edges): Likewise.
7942         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
7943         (duplicate_thunk_for_node): Do not pass frequency.
7944         (cgraph_node::create_clone): Scale only counts.
7945         (cgraph_node::create_virtual_clone): Do not pass frequency.
7946         (cgraph_node::create_edge_including_clones): Do not pass frequency.
7947         (cgraph_node::create_version_clone): Do not pass frequency.
7948         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
7949         (cgraph_node::expand_thunk): Do not pass frequency.
7950         (cgraph_node::create_wrapper): Do not pass frequency.
7951         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
7952         frequency.
7953         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
7954         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
7955         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
7956         (gather_caller_stats): Use frequency function.
7957         (ipcp_cloning_candidate_p): Use frequency function.
7958         (ipcp_propagate_stage): Use frequency function.
7959         (get_info_about_necessary_edges): Use frequency function.
7960         (update_profiling_info): Update only IPA profile.
7961         (update_specialized_profile): Use frequency functoin.
7962         (perhaps_add_new_callers): Update only IPA profile.
7963         * ipa-devirt.c (ipa_devirt): Use IPA profile.
7964         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
7965         (dump_ipa_call_summary): Use frequency function.
7966         (estimate_edge_size_and_time): Use frequency function.
7967         (ipa_merge_fn_summary_after_inlining): Use frequency function.
7968         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
7969         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
7970         (update_noncloned_counts): ... ths one; scale counts only.
7971         (clone_inlined_nodes): Do not scale frequency.
7972         (inline_call): Do not pass frequency.
7973         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
7974         (compute_inlined_call_time): Use IPA profile.
7975         (want_inline_small_function_p): Use IPA profile.
7976         (want_inline_self_recursive_call_p): Use IPA profile.
7977         (edge_badness): Use IPA profile.
7978         (lookup_recursive_calls): Use IPA profile.
7979         (recursive_inlining): Do not pass frequency.
7980         (resolve_noninline_speculation): Do not update frequency.
7981         (inline_small_functions): Collect max of IPA profile.
7982         (dump_overall_stats): Dump IPA porfile.
7983         (dump_inline_stats): Dump IPA porfile.
7984         (ipa_inline): Collect IPA stats.
7985         * ipa-inline.h (clone_inlined_nodes): Update prototype.
7986         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
7987         (ipa_propagate_frequency): Use frequency function.
7988         (ipa_profile): Cleanup.
7989         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
7990         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
7991         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
7992         (input_node): Do not stream frequency.
7993         (input_edge): Do not stream frequency.
7994         (merge_profile_summaries): Scale only IPA profiles.
7995         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
7996         * predict.c (drop_profile): Do not recompute frequency.
7997         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
7998         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
7999         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8000         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8001         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8002         * tree-inline.c (copy_bb): Do not scale frequency.
8003         (expand_call_inline): Do not scale frequency.
8004         (tree_function_versioning): Do not scale frequency.
8005         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8007 2017-11-10  Julia Koval  <julia.koval@intel.com>
8009         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8010         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8011         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8012         (_mm256_maskz_gf2p8affine_epi64_epi8)
8013         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8014         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8015         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8016         (__builtin_ia32_vgf2p8affineqb_v32qi)
8017         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8018         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8020 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8022         PR target/82641
8023         * config/arm/arm.c
8024         (arm_option_override): Refactor.
8025         (arm_option_reconfigure_globals): New.
8026         (arm_options_perform_arch_sanity_checks): New.
8027         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8028         New prototype.
8029         (arm_options_perform_arch_sanity_checks): Likewise
8031 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8033         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8034         (power9-qpmul): New.
8035         * rs6000/rs6000.md ("type" attr): Add qmul.
8036         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8037         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8038         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8040 2017-11-10  Martin Sebor  <msebor@redhat.com>
8042         PR c/81117
8043         * builtins.c (compute_objsize): Handle arrays that
8044         compute_builtin_object_size likes to fail for.  Make extern.
8045         * builtins.h (compute_objsize): Declare.
8046         (check_strncpy_sizes): New function.
8047         (expand_builtin_strncpy): Call check_strncpy_sizes.
8048         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8049         -Wstringop-truncation.
8050         (gimple_fold_builtin_strncat): Same.
8051         * gimple.c (gimple_build_call_from_tree): Set call location.
8052         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8053         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8054         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8055         (handle_builtin_strlen): Use strlen_to_stridx.
8056         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8057         stpncpy.
8058         Use strlen_to_stridx.
8059         (pass_strlen::execute): Release strlen_to_stridx.
8060         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8061         (-Wstringop-truncation): Document new option.
8063 2017-11-10  Martin Liska  <mliska@suse.cz>
8065         PR gcov-profile/82702
8066         * gcov.c (main): Handle intermediate files in a different
8067         way.
8068         (get_gcov_intermediate_filename): New function.
8069         (output_gcov_file): Remove support of intermediate files.
8070         (generate_results): Allocate intermediate file.
8071         (release_structures): Clean-up properly fn_end.
8072         (output_intermediate_file): Start iterating with line 1.
8074 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8076         PR tree-optimization/82929
8077         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8078         ops_swapped_p non-static data member.
8079         (store_immediate_info::store_immediate_info): Clear it.
8080         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8081         ops set ops_swapped_p.
8082         (count_multiple_uses): Handle ops_swapped_p.
8084 2017-11-10  Martin Liska  <mliska@suse.cz>
8086         * coverage.c (coverage_init): Stream information about
8087         support of has_unexecuted_blocks.
8088         * doc/gcov.texi: Document that.
8089         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8090         * gcov.c (read_graph_file): Likewise.
8091         (output_line_beginning): Fix a small issue with
8092         color output.
8094 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8096         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8097         reference of trivial component.
8099 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8101         PR bootstrap/82916
8102         * gimple-ssa-store-merging.c
8103         (pass_store_merging::terminate_all_aliasing_chains): For
8104         gimple_store_p stmts also call refs_output_dependent_p.
8106         PR rtl-optimization/82913
8107         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8108         single set.
8110 2017-11-09  Jeff Law  <law@redhat.com>
8112         * vr-values.h: New file with vr_values class.
8113         * tree-vrp.c: Include vr-values.h
8114         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8115         data objects into the vr_values class.
8116         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8117         (get_value_range): Make it a member function within vr_values class.
8118         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8119         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8120         (op_with_constant_singleton_value_range): Likewise.
8121         (extract_range_for_var_from_comparison_expr): Likewise.
8122         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8123         (extract_range_from_binary_expr): Likewise.
8124         (extract_range_from_unary_expr): Likewise.
8125         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8126         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8127         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8128         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8129         (compare_name_with_value, compare_names): Likewise.
8130         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8131         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8132         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8133         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8134         (extract_range_from_phi_node): Likewise.
8135         (simplify_truth_ops_using_ranges): Likewise.
8136         (simplify_div_or_mod_using_ranges): Likewise.
8137         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8138         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8139         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8140         (simplify_float_conversion_using_ranges): Likewise.
8141         (simplify_internal_call_using_ranges): Likewise.
8142         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8143         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8144         poitner via x_vr_values for calls into gimple folder.
8145         (vrp_initialize_lattice): Make this the vr_values ctor.
8146         (vrp_free_lattice): Make this the vr_values dtor.
8147         (set_vr_value): New function.
8148         (class vrp_prop): Add vr_values data member.  Add various member
8149         functions as well as member functions that delegate to vr_values.
8150         (check_array_ref): Make a member function within vrp_prop class.
8151         (search_for_addr_array, vrp_initialize): Likewise.
8152         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8153         vr_value, values_propagated, etc.
8154         (check_array_bounds): Extract vrp_prop class instance pointer from
8155         walk info structure.  Use it to call member functions.
8156         (check_all_array_refs): Make a member function within vrp_prop class.
8157         Smuggle class instance pointer via walk info structure.
8158         (x_vr_values): New local static.
8159         (vrp_valueize): Use x_vr_values to get class instance.
8160         (vr_valueize_1): Likewise.
8161         (class vrp_folder): Add vr_values data member.  Add various member
8162         functions as well as member functions that delegate to vr_values.
8163         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8164         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8165         class instance from vr_values.  Use it to call member functions.
8166         (vrp_dom_walker): Add vr_values data member.
8167         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8168         instance via x_vr_values.
8169         (identify_jump_threads): Accept vr_values as argument.  Store
8170         it into the walker structure.
8171         (evrp_dom_walker): Add vr_values class data member.  Add various
8172         delegators.
8173         (evrp_dom_walker::try_find_new_range): Use vr_values data
8174         member to access the memory allocator.
8175         (evrp_dom_walker::before_dom_children): Store vr_values class
8176         instance into the vrp_folder class.
8177         (evrp_dom_walker::push_value_range): Rework to avoid direct
8178         access to num_vr_values and vr_value.
8179         (evrp_dom_walker::pop_value_range): Likewise.
8180         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8181         Use vr_values to get to dump_all_value_ranges member function.
8182         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8183         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8184         Pass vr_values class instance down to identify_jump_threads.
8185         Remove call to vrp_free_lattice.
8186         (debug_all_value_ranges): Remove.
8188         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8189         (vrp_folder): Likewise.
8191         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8192         Get it from the existing bitmap instead.
8193         (vrp_intersect_ranges_1): Likewise.
8195 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8197         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8198         bit_not_p field.
8199         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8200         set bit_not_p to it.
8201         (imm_store_chain_info::coalesce_immediate_stores): Break group
8202         if bit_not_p is different.
8203         (count_multiple_uses, split_group,
8204         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8205         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8206         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8207         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8208         to store_immediate_info ctor.
8210 2017-11-09  Jim Wilson  <jimw@sifive.com>
8212         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8213         (scan_prog_file): Likewise.
8215 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8217         * bb-reorder.c (max_entry_frequency): Remove.
8218         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8219         connect_traces, push_to_next_round_p): Remove prototypes.
8220         (find_traces_1_round): Use counts only.
8221         (push_to_next_round_p): Likewise.
8222         (find_traces): Likewise.
8223         (rotate_loop): Likewise.
8224         (find_traces_1_round): Likewise.
8225         (connect_traces): Likewise.
8226         (edge_order): Likewise.
8228 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8230         * config/arm/arm.c (output_return_instruction): Add comments to
8231         indicate requirement for cmse_nonsecure_entry return to account
8232         for the size of clearing instruction output here.
8233         (thumb_exit): Likewise.
8234         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8235         return in hardfloat mode.
8237 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8239         * config/rs6000/rs6000.c (machine_function): Add a bool,
8240         "toc_is_wrapped_separately".
8241         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8242         if it wasn't explicitly set or unset, we are optimizing for speed, and
8243         doing separate shrink-wrapping.
8244         (rs6000_get_separate_components): Enable the TOC component if
8245         saving the TOC register in the prologue.
8246         (rs6000_components_for_bb): Handle the TOC component.
8247         (rs6000_emit_prologue_components): Store the TOC register where needed.
8248         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8249         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8251 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8253         * ipa-param-manipulation.c: New file.
8254         * ipa-param-manipulation.h: Likewise.
8255         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8256         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8257         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8258         (ipa_parm_adjustment): Likewise.
8259         (ipa_parm_adjustment_vec): Likewise.
8260         (ipa_get_vector_of_formal_parms): Moved declaration to
8261         ipa-param-manipulation.h.
8262         (ipa_get_vector_of_formal_parm_types): Likewise.
8263         (ipa_modify_formal_parameters): Likewise.
8264         (ipa_modify_call_arguments): Likewise.
8265         (ipa_combine_adjustments): Likewise.
8266         (ipa_dump_param_adjustments): Likewise.
8267         (ipa_modify_expr): Likewise.
8268         (ipa_get_adjustment_candidate): Likewise.
8269         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8270         ipa-param-manipulation.c.
8271         (ipa_get_vector_of_formal_parm_types): Likewise.
8272         (ipa_modify_formal_parameters): Likewise.
8273         (ipa_modify_call_arguments): Likewise.
8274         (ipa_modify_expr): Likewise.
8275         (get_ssa_base_param): Likewise.
8276         (ipa_get_adjustment_candidate): Likewise.
8277         (index_in_adjustments_multiple_times_p): Likewise.
8278         (ipa_combine_adjustments): Likewise.
8279         (ipa_dump_param_adjustments): Likewise.
8280         * tree-sra.c: Also include ipa-param-manipulation.h
8281         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8282         ipa-param.h.
8284 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8285             Alan Hayward  <alan.hayward@arm.com>
8286             David Sherwood  <david.sherwood@arm.com>
8288         * doc/sourcebuild.texi (vect_masked_store): Document.
8290 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8291             Alan Hayward  <alan.hayward@arm.com>
8292             David Sherwood  <david.sherwood@arm.com>
8294         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8296 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8297             Alan Hayward  <alan.hayward@arm.com>
8298             David Sherwood  <david.sherwood@arm.com>
8300         * doc/sourcebuild.texi (vect_variable_length): Document.
8302 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8303             Alan Hayward  <alan.hayward@arm.com>
8304             David Sherwood  <david.sherwood@arm.com>
8306         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8308 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8309             Alan Hayward  <alan.hayward@arm.com>
8310             David Sherwood  <david.sherwood@arm.com>
8312         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8314 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8315             Alan Hayward  <alan.hayward@arm.com>
8316             David Sherwood  <david.sherwood@arm.com>
8318         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8319         previously undocumented selectors.
8320         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8322 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8324         * doc/rtl.texi (const_vector): Say that elements can be
8325         const_wide_ints too.
8326         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8327         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8328         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8329         * optabs.c (expand_vector_broadcast): Likewise.
8331 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8332             Alan Hayward  <alan.hayward@arm.com>
8333             David Sherwood  <david.sherwood@arm.com>
8335         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8336         scaled index even if the unscaled address was invalid.
8337         Don't increase the complexity of using a scale in that case.
8339 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8340             Alan Hayward  <alan.hayward@arm.com>
8341             David Sherwood  <david.sherwood@arm.com>
8343         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8344         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8345         UNITS_PER_WORD bytes.
8346         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8347         based on REGMODE_NATURAL_SIZE of the inner mode.
8348         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8349         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8350         * expmed.c (lowpart_bit_field_p): Divide the value up into
8351         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8352         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8353         whether something is likely to occupy more than one register.
8355 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8357         PR ipa/82879
8358         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8359         profile_count::adjust_for_ipa_scaling.
8360         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8361         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8362         function.
8363         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8365 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8367         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8368         (split_group): Add total_orig and total_new arguments, estimate the
8369         number of statements related to the store group without store merging
8370         and with store merging.
8371         (imm_store_chain_info::output_merged_store): Adjust split_group
8372         callers, punt if estimated number of statements with store merging
8373         is not smaller than estimated number of statements without it.
8374         Formatting fix.
8375         (handled_load): Remove has_single_use checks.
8376         (pass_store_merging::process_store): Likewise.
8378 2017-11-09  Richard Biener  <rguenther@suse.de>
8380         PR tree-optimization/82902
8381         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8383 2017-11-09  Martin Liska  <mliska@suse.cz>
8385         PR target/82863
8386         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8387         uninitialized.
8389 2017-11-09  Martin Liska  <mliska@suse.cz>
8391         PR tree-optimization/82669
8392         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8394 2017-11-09  Martin Liska  <mliska@suse.cz>
8396         PR gcov-profile/48463
8397         * coverage.c (coverage_begin_function): Output also end locus
8398         of a function and information whether the function is
8399         artificial.
8400         * gcov-dump.c (tag_function): Parse and print the information.
8401         * gcov.c (INCLUDE_MAP): Add include.
8402         (INCLUDE_SET): Likewise.
8403         (struct line_info): Move earlier in the source file because
8404         of vector<line_info> in function_info structure.
8405         (line_info::line_info): Likewise.
8406         (line_info::has_block): Likewise.
8407         (struct source_info): Add new member index.
8408         (source_info::get_functions_at_location): New function.
8409         (function_info::group_line_p): New function.
8410         (output_intermediate_line): New function.
8411         (output_intermediate_file): Use the mentioned function.
8412         (struct function_start): New.
8413         (struct function_start_pair_hash): Likewise.
8414         (process_file): Add code that identifies group functions.
8415         Assign lines either to global or function scope.
8416         (generate_results): Skip artificial functions.
8417         (find_source): Assign index for each source file.
8418         (read_graph_file): Read new flag artificial and end_line.
8419         (add_line_counts): Assign it either to global of function scope.
8420         (accumulate_line_counts): Isolate core of the function to
8421         accumulate_line_info and call it for both function and global
8422         scope lines.
8423         (accumulate_line_info): New function.
8424         (output_line_beginning): Fix GNU coding style.
8425         (print_source_line): New function.
8426         (output_line_details): Likewise.
8427         (output_function_details): Likewise.
8428         (output_lines): Iterate both source (global) scope and function
8429         scope.
8430         (struct function_line_start_cmp): New class.
8431         * doc/gcov.texi: Reflect changes in documentation.
8433 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8435         PR debug/82837
8436         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8437         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8438         and similarly for not instead of neg.
8440 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8442         * config/i386/i386.opt: Add -mforce-indirect-call.
8443         * config/i386/predicates.md: Check for flag_force_indirect_call.
8444         * doc/invoke.texi: Document -mforce-indirect-call
8446 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8448         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8449         New extern.
8450         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8451         config/riscv/riscv.c (predict.h): New include.
8452         (riscv_slow_unaligned_access_p): No longer static.
8453         (riscv_block_move_straight): Add require.
8454         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8456 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8458         PR target/82855
8459         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8460         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8461         nonimmediate_operand predicate for operand 1 instead of
8462         register_operand.
8464 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8466         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8467         New pattern.
8468         * config/aarch64/constraints.md (Uml): New constraint.
8469         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8470         predicate.
8472 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8474         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8475         of two vec_duplicates into a vec_concat.
8477 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8479         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8480         Simplify vec_merge of vec_duplicate and vec_concat.
8481         * config/aarch64/constraints.md (Utq): New constraint.
8482         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8483         define_insn.
8485 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8487         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8488         Simplify vec_merge of vec_duplicate and const_vector.
8489         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8490         New predicate.
8491         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8492         mode iterator.  Update predicate on operand 1 to
8493         handle non-const_vec constants.  Delete constraints.
8494         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8496 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8498         PR tree-optimization/78821
8499         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8500         data member.
8501         (store_operand_info::store_operand_info): Initialize it to false.
8502         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8503         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8504         store in the group, and if chain_info is non-NULL, to ignore altogether
8505         that chain.
8506         (compatible_load_p): Fail if bit_not_p does not match.
8507         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8508         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8509         (pass_store_merging::process_store): Adjust
8510         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8511         call terminate_all_aliasing_chains newly when adding a store into
8512         a chain with non-NULL chain_info.
8514 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8516         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8518 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8520         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8521         Remove.
8522         (aarch64_layout_frame): Initialise emit_frame_chain.
8523         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8524         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8526 2017-11-08  Martin Liska  <mliska@suse.cz>
8528         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8529         of gimple_call_internal_p.
8531 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8533         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8535 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8537         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8538         "do {} while (0)".
8540 2017-11-08  Martin Liska  <mliska@suse.cz>
8542         PR sanitizer/82792
8543         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8545 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8547         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8549 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8551         PR target/82855
8552         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8553         (*cmp<mode>_ccz_1): New insn with $k alternative.
8555         PR target/82855
8556         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8557         RTX_COMM_COMPARE as commutative as well.
8558         (ix86_binary_operator_ok): Formatting fix.
8559         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8560         *<code><mode>3<mask_name><round_saeonly_name>,
8561         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8562         *<s>mul<mode>3_highpart<mask_name>,
8563         *vec_widen_umult_even_v16si<mask_name>,
8564         *vec_widen_umult_even_v8si<mask_name>,
8565         *vec_widen_umult_even_v4si<mask_name>,
8566         *vec_widen_smult_even_v16si<mask_name>,
8567         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8568         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8569         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8570         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8571         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8572         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8573         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8574         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8575         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8576         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8577         ix86_binary_operator_ok.  Formatting fixes.
8578         (*<plusminus_insn><mode>3<mask_name><round_name>,
8579         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8580         fixes.
8582 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8584         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8585         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8586         eq and ne if TARGET_ISEL.
8587         (cmp): New code_iterator.
8588         (UNS, UNSU_, UNSIK): New code_attrs.
8589         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8590         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8591         to...
8592         ("eq<mode>3"): ... this.
8593         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8594         to...
8595         ("ne<mode>3"): ... this.
8597 2017-11-07  Julia Koval  <julia.koval@intel.com>
8599         PR target/82812
8600         * common/config/i386/i386-common.c
8601         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8602         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8603         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8604         * config/i386/i386.opt: Ditto.
8605         * config/i386/i386.c (ix86_target_string): Ditto.
8606         (ix86_option_override_internal): Ditto.
8607         (ix86_init_mpx_builtins): Move MPX to args2.
8608         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8609         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8610         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8611         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8612         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8613         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8614         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8616 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8618         PR target/80425
8619         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8620         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8621         (zero-extendsidi peephole2): Remove peephole.
8623 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8625         PR c/53037
8626         * stor-layout.c: Include attribs.h.
8627         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8628         explicit lookup of "aligned" attribute.
8630 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8632         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8633         (riscv_expand_block_move): Likewise.
8634         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8635         implementation.
8636         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8637         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8638         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8639         (riscv_adjust_block_mem): Likewise.
8640         (riscv_block_move_loop): Likewise.
8641         (riscv_expand_block_move): Likewise.
8642         * config/riscv/riscv.md (movmemsi): New pattern.
8644 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8646         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8647         (MUSL_DYNAMIC_LINKER): Likewise.
8649 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8651         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8652         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8653         nonmmory_operand.
8655 2017-11-07  Richard Biener  <rguenther@suse.de>
8657         * match.pd: Fix build.
8659 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8660             Jackson Woodruff  <jackson.woodruff@arm.com>
8662         PR tree-optimization/71026
8663         * match.pd: Canonicalize negate in division.
8665 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8667         PR middle-end/80131
8668         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8670 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8672         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8673         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8675 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8677         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8678         non-scalar integral types.
8679         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8680         (-(A-B), -(~A)): New transformations.
8682 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8684         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8685         semicolon after "do {} while (0)".
8686         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8687         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8688         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8689         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8690         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8692 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8694         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8695         after "do {} while (0)".
8696         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8697         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8698         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8699         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8700         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8702 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8704         PR other/82784
8705         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8706         "while {} do (0)".
8707         (arm_rtx_costs_internal): Add missing semicolon after
8708         HANDLE_NARROW_SHIFT_ARITH call.
8710 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8712         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8713         disable isel if it was not set explicitly.
8715 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8717         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8718         (add_type_attribute) likewise.
8720 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8722         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8723         of struct ix86_frame.
8724         (ix86_initial_elimination_offset): Likewise.
8725         (ix86_expand_split_stack_prologue): Likewise.
8727 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8729         * tree-vrp.h (enum value_range_type): Update stale comment.
8731 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8732             Alan Hayward  <alan.hayward@arm.com>
8733             David Sherwood  <david.sherwood@arm.com>
8735         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8736         (aarch64_expand_vec_perm_const): Take the number of units too.
8737         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8738         (aarch64_expand_vec_perm_const): Likewise.
8739         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8740         (vec_perm<mode>): Update accordingly.
8742 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8743             Alan Hayward  <alan.hayward@arm.com>
8744             David Sherwood  <david.sherwood@arm.com>
8746         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8747         Take the number of units too.
8748         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8749         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8750         but check for a vector mode before rather than after the call.
8751         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8752         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8753         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8754         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8755         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8756         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8757         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8758         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8759         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8760         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8761         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8762         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8763         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8764         (aarch64_sqdmull2_n<mode>): Update accordingly.
8766 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8767             Alan Hayward  <alan.hayward@arm.com>
8768             David Sherwood  <david.sherwood@arm.com>
8770         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8771         the number of units too.
8772         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8773         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8774         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8775         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8776         (vec_store_lanesxi<mode>): Update accordingly.
8778 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8779             Alan Hayward  <alan.hayward@arm.com>
8780             David Sherwood  <david.sherwood@arm.com>
8782         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8783         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8784         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8785         of units rather than the mode.
8786         * config/aarch64/iterators.md (nunits): New mode attribute.
8787         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8788         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8789         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8790         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8791         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8792         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8793         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8794         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8795         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8796         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8797         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8798         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8799         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8800         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8801         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8802         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8803         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8804         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8805         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8806         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8807         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8808         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8809         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8810         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8811         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8812         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8813         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8814         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8815         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8816         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8817         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8818         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8819         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8820         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8821         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8822         (aarch64_simd_vec_setv2di): Likewise.
8824 2017-11-06  Carl Love  <cel@us.ibm.com>
8826         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8827         definitions.
8828         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8829         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8830         to power 8.
8831         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8832         extern declaration.
8833         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8834         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8835         Add power 8 macro expansions.
8836         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8837         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8838         power 8 instructions.  (VSX_XXBR): Add iterator.
8840 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8842         * config/arm/arm.md (predicable_short_it): Change default to "no",
8843         improve documentation, remove uses that are identical to the default.
8844         (enabled_for_depr_it): Rename to enabled_for_short_it.
8845         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8846         Remove default uses.
8847         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8848         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8849         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8850         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8852 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8854         PR target/82748
8855         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8856         float128 helper macros, which are no longer used after deleting
8857         the old 'q' built-in functions, and moving the round to odd
8858         built-in functions to being special built-in functions.
8859         (BU_FLOAT128_2): Likewise.
8860         (BU_FLOAT128_1_HW): Likewise.
8861         (BU_FLOAT128_2_HW): Likewise.
8862         (BU_FLOAT128_3_HW): Likewise.
8863         (FABSQ): Delete old 'q' built-in functions.
8864         (COPYSIGNQ): Likewise.
8865         (SQRTF128_ODD): Move round to odd built-in functions to be
8866         special built-in functions, so that we can handle
8867         -mabi=ieeelongdouble.
8868         (TRUNCF128_ODD): Likewise.
8869         (ADDF128_ODD): Likewise.
8870         (SUBF128_ODD): Likewise.
8871         (MULF128_ODD): Likewise.
8872         (DIVF128_ODD): Likewise.
8873         (FMAF128_ODD): Likewise.
8874         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
8875         built-in names to 'f128'.
8876         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
8877         old 'q' built-in functions, as the machine independent code for
8878         'f128' built-in functions handles this.
8879         (rs6000_expand_builtin): Add expansion for float128 round to odd
8880         functions, keying off on -mabi=ieeelongdouble of whether to use
8881         the KFmode or TFmode variant.
8882         (rs6000_init_builtins): Initialize the _Float128 round to odd
8883         built-in functions.
8884         * doc/extend.texi (PowerPC Built-in Functions): Document the old
8885         _Float128 'q' built-in functions are now mapped into the new
8886         'f128' built-in functions.
8888 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
8890         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
8891         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
8893 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
8895         PR jit/82826
8896         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
8897         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
8898         * toplev.c: Include "ipa-fnsummary.h".
8899         (toplev::finalize): Call ipa_fnsummary_c_finalize.
8901 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
8903         PR tree-optimization/82838
8904         * gimple-ssa-store-merging.c
8905         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
8906         on a separate gimple_seq which is then appended to seq.
8908 2017-11-06  Jeff Law  <law@redhat.com>
8910         PR target/82788
8911         * config/i386/i386.c (PROBE_INTERVAL): Remove.
8912         (get_probe_interval): New functions.
8913         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
8914         (ix86_adjust_stack_and_probe): Likewise.
8915         (output_adjust_stack_and_probe): Likewise.
8916         (ix86_emit_probe_stack_range): Likewise.
8917         (ix86_expand_prologue): Likewise.
8919 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8921         PR tree-optimization/82816
8922         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
8923         if the modes of the two types are the same.
8924         (convert_plusminus_to_widen): Likewise.
8926 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8928         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
8929         p9_vadu<mode>3.
8930         (usadv16qi): New define_expand.
8931         (usadv8hi): New define_expand.
8933 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
8935         PR bootstrap/82832
8936         * ipa-inline-transform.c (update_noncloned_frequencies): Always
8937         scale.
8938         (inline_transform): Likewise.
8939         * predict.c (counts_to_freqs): Remove useless conditional.
8940         * profile-count.h (profile_count::apply_scale): Move sanity check.
8941         * tree-inline.c (copy_bb): Always scale.
8942         (copy_cfg_body): Likewise.
8944 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
8946         PR target/67591
8947         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
8948         attribute.
8949         (*cmp_ite0): Add enabled_for_depr_it attribute.
8950         (*cmp_ite1): Likewise.
8952 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8954         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
8955         TYPE_MFCRF.
8957 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8959         * tree-vrp.c (vrp_int_const_binop): Return true on success and
8960         return the value by pointer.
8961         (extract_range_from_multiplicative_op_1): Update accordingly.
8962         Return as soon as an operation fails.
8964 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8966         PR other/82784
8967         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
8968         (DEF_SANITIZER_BUILTIN): ... here.
8969         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
8970         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
8972 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8974         PR other/82784
8975         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
8976         macro body.
8977         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
8978         ASM_OUTPUT_BEFORE_CASE_LABEL call.
8979         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
8980         after macro body.
8981         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8982         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8983         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8985 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8987         PR other/82784
8988         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
8989         "do {} while (0)".
8991 2017-11-04  Michael Clark  <michaeljclark@mac.com>
8993         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
8994         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
8995         (adddi3): Likewise.
8996         (*addsi3_extended): Likewise.
8997         (*addsi3_extended2): Likewise.
8998         (<optab>si3): Likewise.
8999         (<optab>di3): Likewise.
9000         (<optab><mode>3): Likewise.
9001         (<*optabe>si3_internal): Likewise.
9002         (zero_extendqi<SUPERQI:mode>2): Likewise.
9003         (*add<mode>hi3): Likewise.
9004         (*xor<mode>hi3): Likewise.
9005         (<optab>di3): Likewise.
9006         (*<optab>si3_extend): Likewise.
9007         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9008         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9009         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9011 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9013         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9014         TARGET_STRICT_ALIGN based upon -mtune argument.
9016 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9018         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9020 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9022         * config/i386/i386.c (choose_basereg): Use optional scratch
9023         register and add assertion.
9024         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9025         needed, and don't allocate stack.
9026         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9027         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9028         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9030 2017-11-03  Jeff Law  <law@redhat.com>
9032         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9033         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9034         to probe at the start of a noreturn function.
9036 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9038         PR tree-optimization/78821
9039         * gimple-ssa-store-merging.c: Update the file comment.
9040         (MAX_STORE_ALIAS_CHECKS): Define.
9041         (struct store_operand_info): New type.
9042         (store_operand_info::store_operand_info): New constructor.
9043         (struct store_immediate_info): Add rhs_code and ops data members.
9044         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9045         and op1r arguments to the ctor, initialize corresponding data members.
9046         (struct merged_store_group): Add load_align_base and load_align
9047         data members.
9048         (merged_store_group::merged_store_group): Initialize them.
9049         (merged_store_group::do_merge): Update them.
9050         (merged_store_group::apply_stores): Pick the constant for
9051         encode_tree_to_bitpos from one of the two operands, or skip
9052         encode_tree_to_bitpos if neither operand is a constant.
9053         (class pass_store_merging): Add process_store method decl.  Remove
9054         bool argument from terminate_all_aliasing_chains method decl.
9055         (pass_store_merging::terminate_all_aliasing_chains): Remove
9056         var_offset_p argument and corresponding handling.
9057         (stmts_may_clobber_ref_p): New function.
9058         (compatible_load_p): New function.
9059         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9060         if there is overlap and rhs_code is not INTEGER_CST.  For
9061         non-overlapping stores terminate group if rhs is not mergeable.
9062         (get_alias_type_for_stmts): Change first argument from
9063         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9064         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9065         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9066         alias type.
9067         (get_location_for_stmts): Change first argument from
9068         auto_vec<gimple *> & to vec<gimple *> &.
9069         (struct split_store): Remove orig_stmts data member, add orig_stores.
9070         (split_store::split_store): Create orig_stores rather than orig_stmts.
9071         (find_constituent_stmts): Renamed to ...
9072         (find_constituent_stores): ... this.  Change second argument from
9073         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9074         to info structures rather than the statements.
9075         (split_group): Rename ALLOW_UNALIGNED argument to
9076         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9077         it.  Adjust find_constituent_stores caller.
9078         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9079         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9080         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9081         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9082         (mem_valid_for_store_merging): New function.
9083         (handled_load): New function.
9084         (pass_store_merging::process_store): New method.
9085         (pass_store_merging::execute): Use process_store method.  Adjust
9086         terminate_all_aliasing_chains caller.
9088 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9090         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9091         Return true for more constants, symbols and label references.
9092         (aarch64_valid_floating_const): Remove unused function.
9094 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9096         PR target/82786
9097         * config/aarch64/aarch64.c (aarch64_layout_frame):
9098         Undo forcing of LR at bottom of frame.
9100 2017-11-03  Jeff Law  <law@redhat.com>
9102         PR target/82823
9103         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9104         for int_registers_saved.
9106         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9107         extracted from tree-ssa-dom.c.
9108         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9109         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9110         (record_equivalences_from_incoming_edge): Add additional argument
9111         to single_pred_edge_ignoring_loop_edges call.
9112         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9113         (uncprop_dom_walker::before_dom_children): Add additional argument
9114         to single_pred_edge_ignoring_loop_edges call.
9115         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9116         single_pred_edge_ignoring_loop_edges rather than open coding.
9117         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9119 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9121         * match.pd (-(-A)): Rewrite.
9123 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9125         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9126         (rs6000_emit_int_cmove): New declaration.
9127         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9128         (rs6000_emit_sISEL): Delete.
9129         (rs6000_emit_int_cmove): Make non-static.
9130         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9131         instead of rs6000_emit_sISEL.
9133 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9135         * asan.c (create_cond_insert_point): Maintain profile.
9136         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9137         merged.
9138         * basic-block.h (struct basic_block_def): Remove frequency.
9139         (EDGE_FREQUENCY): Use to_frequency
9140         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9141         heuristics.
9142         (find_traces): Update to use to_frequency.
9143         (find_traces_1_round): Likewise; use only IPA counts.
9144         (bb_to_key): Likewise.
9145         (connect_traces): Use IPA counts only.
9146         (copy_bb_p): Update to use to_frequency.
9147         (fix_up_crossing_landing_pad): Likewise.
9148         (sanitize_hot_paths): Likewise.
9149         * bt-load.c (basic_block_freq): Likewise.
9150         * cfg.c (init_flow): Set count_max to uninitialized.
9151         (check_bb_profile): Remove frequencies; check counts.
9152         (dump_bb_info): Do not dump frequencies.
9153         (update_bb_profile_for_threading): Update counts only.
9154         (scale_bbs_frequencies_int): Likewise.
9155         (MAX_SAFE_MULTIPLIER): Remove.
9156         (scale_bbs_frequencies_gcov_type): Update counts only.
9157         (scale_bbs_frequencies_profile_count): Update counts only.
9158         (scale_bbs_frequencies): Update counts only.
9159         * cfg.h (struct control_flow_graph): Add count-max.
9160         (update_bb_profile_for_threading): Update prototype.
9161         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9162         (find_many_sub_basic_blocks): Likewise.
9163         * cfgcleanup.c (try_forward_edges): Likewise.
9164         (try_crossjump_to_edge): Likewise.
9165         * cfgexpand.c (expand_gimple_cond): Likewise.
9166         (expand_gimple_tailcall): Likewise.
9167         (construct_init_block): Likewise.
9168         (construct_exit_block): Likewise.
9169         * cfghooks.c (verify_flow_info): Check consistency of counts.
9170         (dump_bb_for_graph): Do not dump frequencies.
9171         (split_block_1): Do not update frequencies.
9172         (split_edge): Do not update frequencies.
9173         (make_forwarder_block): Do not update frequencies.
9174         (duplicate_block): Do not update frequencies.
9175         (account_profile_record): Do not update frequencies.
9176         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9177         for global heuristics.
9178         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9179         (expected_loop_iterations_unbounded): Use counts only.
9180         * cfgloopmanip.c (scale_loop_profile): Simplify.
9181         (create_empty_loop_on_edge): Simplify
9182         (loopify): Simplify
9183         (duplicate_loop_to_header_edge): Simplify
9184         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9185         (update_br_prob_note): Take care of removing note when profile
9186         becomes undefined.
9187         (relink_block_chain): Do not dump frequency.
9188         (rtl_account_profile_record): Use to_frequency.
9189         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9190         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9191         (cgraph_update_edges_for_call_stmt_node): Likewise.
9192         (cgraph_edge::verify_count_and_frequency): Update.
9193         (cgraph_node::verify_node): Temporarily disable frequency verification.
9194         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9195         to_cgraph_frequency.
9196         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9197         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9198         frequencies.
9199         (cgraph_node::expand_thunk): Update profile.
9200         * except.c (dw2_build_landing_pads): Do not update frequency.
9201         * final.c (compute_alignments): Use to_frequency.
9202         (dump_basic_block_info): Do not dump frequency.
9203         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9204         (dump_gimple_bb_header): Do not dump frequency.
9205         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9206         do update count.
9207         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9208         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9209         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9210         (init_before_recovery): Do not update frequency.
9211         (sched_create_recovery_edges): Do not update frequency.
9212         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9213         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9214         (ipa_cp_c_finalize): Set max_count to uninitialized.
9215         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9216         (param_change_prob): Use counts.
9217         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9218         local profiles.
9219         * ipa-split.c (consider_split): Use to_frequency.
9220         (split_function): Use to_frequency.
9221         * ira-build.c (loop_compare_func): Likewise.
9222         (mark_loops_for_removal): Likewise.
9223         (mark_all_loops_for_removal): Likewise.
9224         * loop-doloop.c (doloop_modify): Do not update frequency.
9225         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9226         frequency.
9227         * lto-streamer-in.c (input_function): Update count_max.
9228         * omp-expand.c (expand_omp_taskreg): Update count_max.
9229         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9230         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9231         (maybe_hot_count_p): Use ipa counts only.
9232         (maybe_hot_bb_p): Simplify.
9233         (maybe_hot_edge_p): Simplify.
9234         (probably_never_executed): Do not take frequency argument.
9235         (probably_never_executed_bb_p): Do not pass frequency.
9236         (probably_never_executed_edge_p): Likewise.
9237         (combine_predictions_for_bb): Check that profile is nonzero.
9238         (propagate_freq): Do not set frequency.
9239         (drop_profile): Simplify.
9240         (counts_to_freqs): Simplify.
9241         (expensive_function_p): Use to_frequency.
9242         (propagate_unlikely_bbs_forward): Simplify.
9243         (determine_unlikely_bbs): Simplify.
9244         (estimate_bb_frequencies): Add hack to silence graphite issues.
9245         (compute_function_frequency): Use ipa counts.
9246         (pass_profile::execute): Update.
9247         (rebuild_frequencies): Use counts only.
9248         (force_edge_cold): Use counts only.
9249         * profile-count.c (profile_count::dump): Dump new count types.
9250         (profile_count::differs_from_p): Check compatiblity.
9251         (profile_count::to_frequency): New function.
9252         (profile_count::to_cgraph_frequency): New function.
9253         * profile-count.h (struct function): Declare.
9254         (enum profile_quality): Add profile_guessed_local and
9255         profile_guessed_global0.
9256         (class profile_proability): Decrease number of bits to 29;
9257         update from_reg_br_prob_note and to_reg_br_prob_note.
9258         (class profile_count: Update comment; decrease number of bits
9259         to 61. Check compatibility.
9260         (profile_count::compatible_p): New private member function.
9261         (profile_count::ipa_p): New member function.
9262         (profile_count::operator<): Handle global zero correctly.
9263         (profile_count::operator>): Handle global zero correctly.
9264         (profile_count::operator<=): Handle global zero correctly.
9265         (profile_count::operator>=): Handle global zero correctly.
9266         (profile_count::nonzero_p): New member function.
9267         (profile_count::force_nonzero): New member function.
9268         (profile_count::max): New member function.
9269         (profile_count::apply_scale): Handle IPA scalling.
9270         (profile_count::guessed_local): New member function.
9271         (profile_count::global0): New member function.
9272         (profile_count::ipa): New member function.
9273         (profile_count::to_frequency): Declare.
9274         (profile_count::to_cgraph_frequency): Declare.
9275         * profile.c (OVERLAP_BASE): Delete.
9276         (compute_frequency_overlap): Delete.
9277         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9278         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9279         * sched-ebb.c (rank): Use counts only.
9280         * shrink-wrap.c (handle_simple_exit): Use counts only.
9281         (try_shrink_wrapping): Use counts only.
9282         (place_prologue_for_one_component): Use counts only.
9283         * tracer.c (find_best_predecessor): Use to_frequency.
9284         (find_trace): Use to_frequency.
9285         (tail_duplicate): Use to_frequency.
9286         * trans-mem.c (expand_transaction): Do not update frequency.
9287         * tree-call-cdce.c: Do not update frequency.
9288         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9289         (gimple_merge_blocks): Likewise.
9290         (gimple_split_edge): Likewise.
9291         (gimple_duplicate_sese_region): Likewise.
9292         (gimple_duplicate_sese_tail): Likewise.
9293         (move_sese_region_to_fn): Likewise.
9294         (gimple_account_profile_record): Likewise.
9295         (insert_cond_bb): Likewise.
9296         * tree-complex.c (expand_complex_div_wide): Likewise.
9297         * tree-eh.c (lower_resx): Update profile.
9298         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9299         frequencies.
9300         (initialize_cfun): Do not initialize frequencies
9301         (freqs_to_counts): Delete.
9302         (copy_cfg_body): Ignore count parameter.
9303         (copy_body): Update.
9304         (expand_call_inline): Update count_max.
9305         (optimize_inline_calls): Update count_max.
9306         (tree_function_versioning): Update count_max.
9307         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9308         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9309         frequency.
9310         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9311         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9312         (try_peel_loop): Likewise.
9313         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9314         to_frequency.
9315         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9316         (tree_transform_and_unroll_loop): Do not use frequencies
9317         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9318         Use reliable prediction only.
9319         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9320         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9321         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9322         probability scaling.
9323         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9324         not update frequency
9325         (any_remaining_duplicated_blocks): Likewise.
9326         (update_profile): Likewise.
9327         (estimated_freqs_path): Delete.
9328         (freqs_to_counts_path): Delete.
9329         (clear_counts_path): Delete.
9330         (ssa_fix_duplicate_block_edges): Likewise.
9331         (duplicate_thread_path): Likewise.
9332         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9333         * tree-tailcall.c (decrease_profile): Do not update frequency.
9334         (eliminate_tail_call): Likewise.
9335         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9336         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9337         (optimize_mask_stores): Likewise.
9338         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9339         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9340         (ubsan_expand_ptr_ifn): Update profile.
9341         * value-prof.c (gimple_ic): Simplify.
9342         * value-prof.h (gimple_ic): Update prototype.
9343         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9344         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9345         counts are nonzero.
9346         (want_inline_self_recursive_call_p): Likewise.
9347         (resolve_noninline_speculation): Only cummulate defined counts.
9348         (inline_small_functions): Use nonzero_p.
9349         (ipa_inline): Do not access freed node.
9351 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9353         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9354         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9356 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9358         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9359         non-legitimate address.
9361 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9363         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9364         (*lt0_<mode>di, *lt0_<mode>si): New.
9366 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9368         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9369         TARGET_PAIRED_FLOAT.
9371 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9372             Jim Wilson  <jim.wilson@linaro.org>
9374         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9375         * config/aarch64/aarch64-tune.md: Regenerated.
9376         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9377         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9379 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9381         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9382         (arc_expand_prologue): Restore blink for millicode.
9383         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9385 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9387         PR target/82809
9388         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9389         gen_vec_duplicate after forcing the scalar into a register.
9391 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9393         * combine (try_combine): Print the insns input to try_combine to the
9394         dump file.
9396 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9398         PR target/79868
9399         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9400         Remove second argument from aarch64_process_target_attr call.
9401         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9402         Ditto.
9403         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9404         field type.
9405         (aarch64_handle_attr_arch): Remove second argument.
9406         (aarch64_handle_attr_cpu): Ditto.
9407         (aarch64_handle_attr_tune): Ditto.
9408         (aarch64_handle_attr_isa_flags): Ditto.
9409         (aarch64_process_one_target_attr): Ditto.
9410         (aarch64_process_target_attr): Ditto.
9411         (aarch64_option_valid_attribute_p): Remove second argument.
9412         on aarch64_process_target_attr call.
9414 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9416         * diagnostic.c: Include "selftest-diagnostic.h".
9417         (selftest::assert_location_text): New function.
9418         (selftest::test_diagnostic_get_location_text): New function.
9419         (selftest::diagnostic_c_tests): Call it.
9421 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9423         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9424         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9425         (class selftest::test_diagnostic_context): Move to...
9426         * selftest-diagnostic.c: New file.
9427         * selftest-diagnostic.h: New file.
9429 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9431         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9432         offset range for FT32B.
9433         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9434         * config/ft32/ft32.md: Add TARGET_NOPM.
9435         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9436         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9438 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9440         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9442 2017-11-02  Jeff Law  <law@redhat.com>
9444         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9445         virtual keyword on FINAL OVERRIDE members.
9447         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9448         virtuals together.  Add virtual destructor.
9449         (substitute_and_fold_engine): Similarly.
9451 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9453         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9455 2017-11-02  Richard Biener  <rguenther@suse.de>
9457         PR tree-optimization/82795
9458         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9460 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9462         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9463         gcc_SUN_LD_VERSION.
9464         (gcc_GAS_CHECK_FEATURE): Remove.
9465         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9466         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9467         * configure: Regenerate.
9469 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9471         * config/arc/arc.c (hwloop_optimize): Account for empty
9472         body loops.
9474 2017-11-02  Richard Biener  <rguenther@suse.de>
9476         PR middle-end/82765
9477         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9478         Truncate ARRAY_REF index and element size.
9480 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9482         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9484 2017-11-01  Jeff Law  <law@redhat.com>
9486         * tree-ssa-ccp.c (ccp_folder): New class derived from
9487         substitute_and_fold_engine.
9488         (ccp_folder::get_value): New member function.
9489         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9490         (ccp_fold_stmt): Remove prototype.
9491         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9492         * tree-ssa-copy.c (copy_folder): New class derived from
9493         substitute_and_fold_engine.
9494         (copy_folder::get_value): Renamed from get_value.
9495         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9496         * tree-vrp.c (vrp_folder): New class derived from
9497         substitute_and_fold_engine.
9498         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9499         (vrp_folder::get_value): New member function.
9500         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9501         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9502         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9503         provide a class interface to folder/substitute routines.
9504         (ssa_prop_fold_stmt_fn): Remove typedef.
9505         (ssa_prop_get_value_fn): Likewise.
9506         (subsitute_and_fold): Remove prototype.
9507         (replace_uses_in): Likewise.
9508         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9509         Renamed from replace_uses_in.  Call the virtual member function
9510         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9511         (substitute_and_fold_dom_walker): Remove initialization of
9512         data member entries for calbacks.  Add substitute_and_fold_engine
9513         member and initialize it.
9514         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9515         member functions for get_value, replace_phi_args_in c
9516         replace_uses_in, and fold_stmt calls.
9517         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9518         substitute_and_fold.  Remove assert.   Update ctor call.
9520         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9521         (ssa_prop_visit_phi_fn): Likewise.
9522         (class ssa_propagation_engine): New class to provide an interface
9523         into ssa_propagate.
9524         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9525         variable.
9526         (ssa_prop_visit_phi): Likewise.
9527         (ssa_propagation_engine::simulate_stmt): Moved into class.
9528         Call visit_phi/visit_stmt from the class rather than via
9529         file scoped static variables.
9530         (ssa_propagation_engine::simulate_block): Moved into class.
9531         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9532         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9533         set file scoped statics for the visit_stmt/visit_phi callbacks.
9534         * tree-complex.c (complex_propagate): New class derived from
9535         ssa_propagation_engine.
9536         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9537         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9538         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9539         class.
9540         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9541         ssa_propagation_engine.
9542         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9543         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9544         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9545         * tree-ssa-copy.c (copy_prop): New class derived from
9546         ssa_propagation_engine.
9547         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9548         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9549         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9550         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9551         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9552         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9553         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9555 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9557         PR rtl-optimization/82778
9558         PR rtl-optimization/82597
9559         * compare-elim.c (struct comparison): Add in_a_setter field.
9560         (find_comparison_dom_walker::before_dom_children): Remove killed
9561         bitmap and df_simulate_find_defs call, instead walk the defs.
9562         Compute last_setter and initialize in_a_setter.  Merge definitions
9563         with first initialization for a few variables.
9564         (try_validate_parallel): Use insn_invalid_p instead of
9565         recog_memoized.  Return insn rather than just the pattern.
9566         (try_merge_compare): Fix up comment.  Don't uselessly test if
9567         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9568         chains.
9569         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9570         call.
9572 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9573             Alan Hayward  <alan.hayward@arm.com>
9574             David Sherwood  <david.sherwood@arm.com>
9576         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9577         aarch64_hard_regno_nregs to get the number of registers
9578         in a mode.
9580 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9581             Alan Hayward  <alan.hayward@arm.com>
9582             David Sherwood  <david.sherwood@arm.com>
9584         * config/aarch64/constraints.md (Upl): Rename to...
9585         (Uaa): ...this.
9586         * config/aarch64/aarch64.md
9587         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9588         Update accordingly.
9590 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9591             Alan Hayward  <alan.hayward@arm.com>
9592             David Sherwood  <david.sherwood@arm.com>
9594         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9595         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9596         earlier in file.
9598 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9599             Alan Hayward  <alan.hayward@arm.com>
9600             David Sherwood  <david.sherwood@arm.com>
9602         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9603         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9604         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9605         named expanders.
9606         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9607         of a single element.
9608         * config/aarch64/iterators.md: Add a comment above the permute
9609         unspecs to say that they are generated directly by
9610         aarch64_expand_vec_perm_const.
9611         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9613 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9615         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9617 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9619         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9620         and medany code models, and describe what they do.
9622 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9624         Revert accidental duplicate:
9626         * combine.c (can_change_dest_mode): Reject changes in
9627         REGMODE_NATURAL_SIZE.
9629 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9631         PR rtl-optimization/64682
9632         PR rtl-optimization/69567
9633         PR rtl-optimization/69737
9634         PR rtl-optimization/82683
9635         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9636         register mentioned in the note, drop the note, unless it came from I3,
9637         in which case it should go to I3 again.
9639 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9641         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9642         and return false if not.
9643         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9645 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9647         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9648         range is known to be empty.
9650 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9651             Alan Hayward  <alan.hayward@arm.com>
9652             David Sherwood  <david.sherwood@arm.com>
9654         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9655         and CONST_VECTOR_NUNITS instead of computing the number of units from
9656         the byte sizes of the vector and element.
9657         (simplify_binary_operation_1): Likewise.
9658         (simplify_const_binary_operation): Likewise.
9659         (simplify_ternary_operation): Likewise.
9661 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9662             Alan Hayward  <alan.hayward@arm.com>
9663             David Sherwood  <david.sherwood@arm.com>
9665         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9666         (int_mem_offset): ...this new function.
9667         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9668         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9669         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9670         Update accordingly.
9672 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9673             Alan Hayward  <alan.hayward@arm.com>
9674             David Sherwood  <david.sherwood@arm.com>
9676         * lower-subreg.c (interesting_mode_p): New function.
9677         (compute_costs, find_decomposable_subregs, decompose_register)
9678         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9679         (resolve_clobber, dump_choices): Use it.
9681 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9682             Alan Hayward  <alan.hayward@arm.com>
9683             David Sherwood  <david.sherwood@arm.com>
9685         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9687 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9688             Alan Hayward  <alan.hayward@arm.com>
9689             David Sherwood  <david.sherwood@arm.com>
9691         * alias.c (find_base_value, find_base_term): Only process integer
9692         truncations.  Check the precision rather than the size.
9694 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9695             Alan Hayward  <alan.hayward@arm.com>
9696             David Sherwood  <david.sherwood@arm.com>
9698         * machmode.h (is_narrower_int_mode): New function
9699         * optabs.c (expand_float, expand_fix): Use it.
9700         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9702 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9703             Alan Hayward  <alan.hayward@arm.com>
9704             David Sherwood  <david.sherwood@arm.com>
9706         * rtl.h (narrower_subreg_mode): New function.
9707         * ira-color.c (update_costs_from_allocno): Use it.
9709 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9710             Alan Hayward  <alan.hayward@arm.com>
9711             David Sherwood  <david.sherwood@arm.com>
9713         * optabs-query.h (convert_optab_p): New function, split out from...
9714         (convert_optab_handler): ...here.
9715         (widening_optab_handler): Delete.
9716         (find_widening_optab_handler): Remove permit_non_widening parameter.
9717         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9718         override that operates on mode class wrappers.
9719         * optabs-query.c (widening_optab_handler): Delete.
9720         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9721         parameter.  Assert that the two modes are the same class and that
9722         the "from" mode is narrower than the "to" mode.  Use
9723         convert_optab_handler instead of widening_optab_handler.
9724         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9725         instead of widening_optab_handler.
9726         * expr.c (expand_expr_real_2): Update calls to
9727         find_widening_optab_handler.
9728         * optabs.c (expand_widen_pattern_expr): Likewise.
9729         (expand_binop_directly): Take the insn_code as a parameter.
9730         (expand_binop): Only call find_widening_optab_handler for
9731         conversion optabs; use optab_handler otherwise.  Update calls
9732         to find_widening_optab_handler and expand_binop_directly.
9733         Use convert_optab_handler instead of widening_optab_handler.
9734         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9735         find_widening_optab_handler and use scalar_mode rather than
9736         machine_mode.
9737         (convert_plusminus_to_widen): Likewise.
9739 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9740             Alan Hayward  <alan.hayward@arm.com>
9741             David Sherwood  <david.sherwood@arm.com>
9743         * machmode.h (fixed_size_mode): New class.
9744         * rtl.h (get_pool_mode): Return fixed_size_mode.
9745         * gengtype.c (main): Add fixed_size_mode.
9746         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9747         (get_raw_arg_mode): Likewise.
9748         * doc/tm.texi: Regenerate.
9749         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9750         * targhooks.c (default_get_reg_raw_mode): Likewise.
9751         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9752         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9753         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9754         (msp430_get_raw_result_mode): Likewise.
9755         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9756         * dbxout.c (dbxout_parms): Require fixed-size modes.
9757         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9758         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9759         * omp-low.c (lower_oacc_reductions): Likewise.
9760         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9761         (simplify_subreg): Update accordingly.
9762         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9763         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9764         that aren't fixed-size.
9765         (get_pool_mode): Return a fixed_size_mode.
9766         (output_constant_pool_2): Take a fixed_size_mode.
9768 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9769             Alan Hayward  <alan.hayward@arm.com>
9770             David Sherwood  <david.sherwood@arm.com>
9772         * doc/rtl.texi (vec_series): Document.
9773         (const): Say that the operand can be a vec_series.
9774         * rtl.def (VEC_SERIES): New rtx code.
9775         * rtl.h (const_vec_series_p_1): Declare.
9776         (const_vec_series_p): New function.
9777         * emit-rtl.h (gen_const_vec_series): Declare.
9778         (gen_vec_series): Likewise.
9779         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9780         (gen_vec_series): Likewise.
9781         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9782         * simplify-rtx.c (simplify_unary_operation): Handle negations
9783         of vector series.
9784         (simplify_binary_operation_series): New function.
9785         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9786         (test_vector_ops_series): New function.
9787         (test_vector_ops): Call it.
9788         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9789         gen_const_vec_series.
9790         (altivec_lvsr): Likewise.
9791         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9793 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9794             Alan Hayward  <alan.hayward@arm.com>
9795             David Sherwood  <david.sherwood@arm.com>
9797         * doc/rtl.texi (const): Update description of address constants.
9798         Say that vector constants are allowed too.
9799         * common.md (E, F): Use CONSTANT_P instead of checking for
9800         CONST_VECTOR.
9801         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9802         checking for CONST_VECTOR.
9803         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9804         VEC_DUPLICATE.
9805         * expr.c (expand_expr_real_2): Check for vector modes instead
9806         of checking for CONST_VECTOR.
9807         * rtl.h (const_vec_p): New function.
9808         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9809         (unwrap_const_vec_duplicate): Handle them here too.
9811 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9812             David Malcolm  <dmalcolm@redhat.com>
9813             Alan Hayward  <alan.hayward@arm.com>
9814             David Sherwood  <david.sherwood@arm.com>
9816         * rtl.h (vec_duplicate_p): New function.
9817         * selftest-rtl.c (assert_rtx_eq_at): New function.
9818         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9819         (assert_rtx_eq_at): Declare.
9820         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9821         * selftest-run-tests.c (selftest::run_tests): Call it.
9822         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9823         (simplify_unary_operation_1): Recursively handle vector duplicates.
9824         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9825         vector duplicates.
9826         (simplify_subreg): Handle subregs of vector duplicates.
9827         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9828         (selftest::simplify_rtx_c_tests): New functions.
9830 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9831             Alan Hayward  <alan.hayward@arm.com>
9832             David Sherwood  <david.sherwood@arm.com>
9834         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9835         (gen_vec_duplicate): Likewise.
9836         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9837         out from...
9838         (gen_const_vector): ...here.
9839         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9840         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9841         whose elements are all equal.
9842         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9843         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9844         (simplify_relational_operation): Likewise.
9845         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9846         Likewise.
9847         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9848         (aarch64_expand_vector_init): Likewise.
9849         * config/arm/arm.c (neon_vdup_constant): Likewise.
9850         (neon_expand_vector_init): Likewise.
9851         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9852         (arm_block_set_unaligned_vect): Likewise.
9853         (arm_block_set_aligned_vect): Likewise.
9854         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9855         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9856         (expand_vec_perm_even_odd_pack): Likewise.
9857         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9858         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9859         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9860         gen_const_vec_duplicate.
9861         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9862         * config/mips/mips.c (mips_gen_const_int_vector): Use
9863         gen_const_vec_duplicate.
9864         (mips_expand_vector_init): Use CONST0_RTX.
9865         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
9866         (define_split): Use gen_const_vec_duplicate.
9867         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
9868         (define_split): Use gen_const_vec_duplicate.
9869         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
9870         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
9871         * config/spu/spu.c (spu_const): Likewise.
9873 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9874             Alan Hayward  <alan.hayward@arm.com>
9875             David Sherwood  <david.sherwood@arm.com>
9877         * combine.c (can_change_dest_mode): Reject changes in
9878         REGMODE_NATURAL_SIZE.
9880 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
9882         * configure.ac (--enable-libssp): New.
9883         (gcc_cv_libc_provides_ssp): Check for explicit setting before
9884         trying to determine target-specific default.  Adjust indentation.
9885         * configure: Regenerated.
9886         * doc/install.texi (Configuration): Expand --disable-libssp
9887         documentation.
9889 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
9891         config/i386/i386.c (ix86_expand_epilogue): Correct stack
9892         calculation.
9894 2017-10-31  Martin Jambor  <mjambor@suse.cz>
9896         PR c++/81702
9897         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
9899 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
9901         * auto-profile.c (autofdo_source_profile::read): Use
9902         UNKNOWN_LOCATION rather than 0.
9903         * diagnostic-core.h (warning_at_rich_loc): Rename to...
9904         (warning_at): ...this overload.
9905         (warning_at_rich_loc_n): Rename to...
9906         (warning_n): ...this overload.
9907         (error_at_rich_loc): Rename to...
9908         (error_at): ...this overload.
9909         (pedwarn_at_rich_loc): Rename to...
9910         (pedwarn): ...this overload.
9911         (permerror_at_rich_loc): Rename to...
9912         (permerror): ...this overload.
9913         (inform_at_rich_loc): Rename to...
9914         (inform): ...this overload.
9915         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
9916         (diagnostic_n_impl_richloc): Rename to...
9917         (diagnostic_n_impl): ...this rich_location *-based decl.
9918         (inform_at_rich_loc): Rename to...
9919         (inform): ...this, and add an assertion.
9920         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
9921         (warning_at_rich_loc): Rename to...
9922         (warning_at): ...this, and add an assertion.
9923         (warning_at_rich_loc_n): Rename to...
9924         (warning_n): ...this, and add an assertion.
9925         (warning_n): Update location_t-based implementation for removal of
9926         location_t-based diagnostic_n_impl.
9927         (pedwarn_at_rich_loc): Rename to...
9928         (pedwarn): ...this, and add an assertion.
9929         (permerror_at_rich_loc): Rename to...
9930         (permerror): ...this, and add an assertion.
9931         (error_n): Update for removal of location_t-based diagnostic_n_impl.
9932         (error_at_rich_loc): Rename to...
9933         (error_at): ...this, and add an assertion.
9934         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
9935         (driver::do_spec_on_infiles): Likewise.
9936         * substring-locations.c (format_warning_va): Update for renaming
9937         of inform_at_rich_loc.
9939 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
9941         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
9942         _Float<N>X built-in functions so that the variant without the
9943         "__builtin_" prefix is only enabled for the GNU C and Objective C
9944         languages when they are in non-strict ANSI/ISO mode.
9945         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
9946         * target.def (floatn_builtin_p): Add a target hook to control
9947         whether _Float<N> and _Float<N>X built-in functions without the
9948         "__builtin_" prefix are enabled, and return true for C and
9949         Objective C in the default hook.  Include langhooks.h in
9950         targhooks.c.
9951         * targhooks.h (default_floatn_builtin_p): Likewise.
9952         * targhooks.c (default_floatn_builtin_p): Likewise.
9953         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
9954         floatn_builtin_p target hook.
9955         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
9957 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
9958             Eric Botcazou  <ebotcazou@adacore.com>
9960         PR rtl-optimization/81803
9961         * lra-constraints.c (curr_insn_transform): Also reload the whole
9962         register for a strict subreg no wider than a word if this is for
9963         a WORD_REGISTER_OPERATIONS target.
9965 2017-10-31  Jason Merrill  <jason@redhat.com>
9967         * gdbinit.in: Skip over inlines from timevar.h.
9969 2017-10-31  Martin Liska  <mliska@suse.cz>
9971         * doc/gcov.texi: Document new option.
9972         * gcov.c (print_usage): Likewise print it.
9973         (process_args): Support the argument.
9974         (format_count): New function.
9975         (format_gcov): Use the function.
9977 2017-10-31  Martin Liska  <mliska@suse.cz>
9979         * gcov.c (struct name_map): do not use typedef.
9980         Define operator== and operator<.
9981         (name_search): Remove.
9982         (name_sort): Remove.
9983         (main): Do not allocate names.
9984         (process_file): Add vertical space.
9985         (generate_results): Use std::find.
9986         (release_structures): Do not release memory.
9987         (find_source): Use std::find.
9989 2017-10-31  Martin Liska  <mliska@suse.cz>
9991         * gcov.c (struct line_info): Remove it's typedef.
9992         (line_info::line_info): Add proper ctor.
9993         (line_info::has_block): Do not use a typedef.
9994         (struct source_info): Do not use typedef.
9995         (circuit): Likewise.
9996         (get_cycles_count): Likewise.
9997         (output_intermediate_file): Iterate via vector iterator.
9998         (add_line_counts): Use std::vector methods.
9999         (accumulate_line_counts): Likewise.
10000         (output_lines): Likewise.
10002 2017-10-31  Martin Liska  <mliska@suse.cz>
10004         * gcov.c (struct source_info): Remove typedef.
10005         (source_info::source_info): Add proper ctor.
10006         (accumulate_line_counts): Use struct, not it's typedef.
10007         (output_gcov_file): Likewise.
10008         (output_lines): Likewise.
10009         (main): Do not allocate an array.
10010         (output_intermediate_file): Use size of vector container.
10011         (process_file): Resize the vector.
10012         (generate_results): Do not preallocate, use newly added vector
10013         lines.
10014         (release_structures): Do not release sources.
10015         (find_source): Use vector methods.
10016         (add_line_counts): Do not use typedef.
10018 2017-10-31  Martin Liska  <mliska@suse.cz>
10020         * doc/gcov.texi: Document that.
10021         * gcov.c (add_line_counts): Mark lines with a non-executed
10022         statement.
10023         (output_line_beginning): Handle such lines.
10024         (output_lines): Pass new argument.
10025         (output_intermediate_file): Print it in intermediate format.
10027 2017-10-31  Martin Liska  <mliska@suse.cz>
10029         * color-macros.h: New file.
10030         * diagnostic-color.c: Factor out color related to macros to
10031         color-macros.h.
10032         * doc/gcov.texi: Document -k option.
10033         * gcov.c (INCLUDE_STRING): Include string.h.
10034         (print_usage): Add -k option.
10035         (process_args): Parse it.
10036         (pad_count_string): New function.
10037         (output_line_beginning): Likewise.
10038         (DEFAULT_LINE_START): New macro.
10039         (output_lines): Support color output.
10041 2017-10-31  Martin Liska  <mliska@suse.cz>
10043         PR gcov-profile/82633
10044         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10045         their interaction with GCOV infrastructure.
10046         * configure.ac: Add -fkeep-{inline,static}-functions to
10047         coverage_flags.
10048         * configure: Regenerate.
10050 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10052         PR target/82772
10053         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10055 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10057         PR target/82674
10058         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10059         into a register if it does not fit into an immediate offset field.
10061 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10063         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10065 2017-10-31  Julia Koval  <julia.koval@intel.com>
10067         * config.gcc: Add gfniintrin.h.
10068         * config/i386/gfniintrin.h: New.
10069         * config/i386/i386-builtin-types.def
10070         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10071         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10072         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10073         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10074         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10075         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10076         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10077         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10078         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10079         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10080         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10081         * config/i386/immintrin.h: Include gfniintrin.h.
10082         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10084 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10086         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10088 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10090         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10091         (arm_ashldi3_1bit): Remove pattern.
10092         (ashrdi3): Remove shift by 1 expansion.
10093         (arm_ashrdi3_1bit): Remove pattern.
10094         (lshrdi3): Remove shift by 1 expansion.
10095         (arm_lshrdi3_1bit): Remove pattern.
10096         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10097         cost of ashldi3 by 1.
10098         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10099         (<shift>di3_neon): Likewise.
10101 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10103         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10104         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10105         and (*aarch64_simd_mov<VQ:mode>).
10106         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10107         pattern alternative.
10108         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10109         attributes to match pattern alternative.
10111 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10113         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10114         * config/rs6000/emmintrin.h: New file.
10115         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10117 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10119         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10120         * (movdi_vfp_cortexa8): Remove pattern.
10122 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10124         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10125         etc. reference.
10126         (Specific, alpha*-dec-osf5.1): Remove.
10127         (Specific, mips-sgi-irix5): Remove.
10128         (Specific, mips-sgi-irix6): Remove.
10130 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10132         PR middle-end/22141
10133         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10134         arguments to clear_bit_region_be.
10136 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10138         * gimplify.c: Include memmodel.h.
10140 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10142         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10143         condition whether loop should be executed at all.
10145 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10147         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10148         gimple folding of vec_madd() intrinsics.
10149         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10150         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10151         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10153 2017-10-30  Richard Biener  <rguenther@suse.de>
10155         PR tree-optimization/82762
10156         Revert
10157         2017-10-23  Richard Biener  <rguenther@suse.de>
10159         PR tree-optimization/82129
10160         Revert
10161         2017-08-01  Richard Biener  <rguenther@suse.de>
10163         PR tree-optimization/81181
10164         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10165         (compute_antic): ... end of iteration here.
10167 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10169         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10170         -std=iso9899:2017 and -std=gnu17.
10171         * doc/standards.texi (C Language): Document C17 support.
10172         * doc/cpp.texi (Overview): Mention -std=c17.
10173         (Standard Predefined Macros): Document C11 and C17 values of
10174         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10175         * doc/extend.texi (Inline): Do not list individual options for
10176         standards newer than C99.
10177         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10178         "GNU C17".
10179         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10180         language name.
10182 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10184         * asan.c (asan_finish_file): Align asan globals array by shadow
10185         granularity.
10187 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10189         PR middle-end/22141
10190         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10191         (struct store_immediate_info): Add bitregion_start and bitregion_end
10192         fields.
10193         (store_immediate_info::store_immediate_info): Add brs and bre
10194         arguments and initialize bitregion_{start,end} from those.
10195         (struct merged_store_group): Add bitregion_start, bitregion_end,
10196         align_base and mask fields.  Drop unnecessary struct keyword from
10197         struct store_immediate_info.  Add do_merge method.
10198         (clear_bit_region_be): Use memset instead of loop storing zeros.
10199         (merged_store_group::do_merge): New method.
10200         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10201         stores as long as the surrounding bitregions have no gaps.
10202         (merged_store_group::merge_overlapping): Use do_merge.
10203         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10204         is byte aligned, rather than requiring that start and width are
10205         byte aligned.  Drop unnecessary struct keyword from
10206         struct store_immediate_info.  Allocate and populate also mask array.
10207         Make start of the arrays relative to bitregion_start rather than
10208         start and size them according to bitregion_{end,start} difference.
10209         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10210         struct store_immediate_info.
10211         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10212         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10213         struct keyword from struct store_immediate_info.
10214         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10215         between stores as long as the surrounding bitregions have no gaps.
10216         Formatting fixes.
10217         (struct split_store): Add orig non-static data member.
10218         (split_store::split_store): Initialize orig to false.
10219         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10220         if there is exactly a single original stmt.  Change stmts argument
10221         to pointer from reference, if NULL, don't push anything to it.  Add
10222         first argument, use it to optimize skipping over orig stmts that
10223         are known to be before bitpos already.  Simplify.
10224         (split_group): Return unsigned int count how many stores are or
10225         would be needed rather than a bool.  Add allow_unaligned argument.
10226         Change split_stores argument from reference to pointer, if NULL,
10227         only do a dry run computing how many stores would be produced.
10228         Rewritten algorithm to use both alignment and misalign if
10229         !allow_unaligned and handle bitfield stores with gaps.
10230         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10231         from bitregion_start instead of start.  Compute allow_unaligned
10232         here, if true, do 2 split_group dry runs to compute which one
10233         produces fewer stores and prefer aligned if equal.  Punt if
10234         new count is bigger or equal than original before emitting any
10235         statements, rather than during that.  Remove no longer needed
10236         new_ssa_names tracking.  Replace num_stmts with
10237         split_stores.length ().  Use 32-bit stack allocated entries
10238         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10239         if possible.  Handle bitfields with gaps.
10240         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10241         Compute bitregion_{start,end} for the stores and construct
10242         store_immediate_info with that.  Formatting fixes.
10244 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10246         PR target/82725
10247         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10248         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10250 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10252         * gimplify.c: Include tm_p.h.
10254         * common.opt (gcoff): Re-add as ignored option.
10255         (gcoff1, gcoff2, gcoff3): Likewise.
10257         * Makefile.in (OBJS): Delete sdbout.o.
10258         (GTFILES): Delete $(srcdir)/sdbout.c.
10259         * debug.h: Delete sdb_debug_hooks.
10260         * final.c: Delete sdbout.h include.
10261         (final_scan_insn): Delete SDB_DEBUG check.
10262         (rest_of_clean_state): Likewise.
10263         * output.h: Delete sdb_begin_function_line.
10264         * sdbout.c: Delete.
10265         * sdbout.h: Delete.
10266         * toplev.c: Delete sdbout.h include.
10267         (process_options): Delete SDB_DEBUG check.
10268         * tree-core.h (tree_type_common): Delete pointer field of
10269         tree_type_symtab.
10270         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10271         TYPE_SYMTAB_POINTER.
10272         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10273         (TYPE_SYMTAB_IS_POINTER): Delete.
10274         (TYPE_SYMTAB_IS_DIE): Renumber.
10275         * xcoffout.c: Refer to former sdbout.c file.
10276         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10278         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10279         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10280         (Debugging Options): Delete -gcoff.
10281         (-gstabs): Delete SDB reference.
10282         (-gcoff): Delete.
10283         (-gcoff@var{level}): Delete.
10284         * doc/passes.texi (Debugging information output): Delete SDB and
10285         sdbout.c references.
10286         * doc/tm.texi: Regenerate.
10287         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10288         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10289         references.
10290         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10291         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10292         (SDB_DEBUGGING_INFO): Delete.
10293         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10294         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10295         * target.def (output_source_filename): Delete COFF reference.
10297         * common.opt (gcoff): Delete.
10298         (gxcoff+): Update Negative chain.
10299         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10300         SDB_DEBUG.
10301         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10302         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10303         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10304         and SDB references.
10305         (expand_function_start): Change sdb reference to past tense.
10306         (expand_function_end): Change sdb reference to past tense.
10307         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10308         * opts.c (debug_type_names): Delete coff entry.
10309         (common_handle_option): Delete OPT_gcoff case.
10310         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10312         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10313         * config/cris/cris.h: Delete SDB reference in comment.
10314         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10315         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10316         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10317         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10318         to past tense.
10319         (ix86_expand_prologue): Likewise.
10320         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10321         * config/ia64/ia64.h: Likewise.
10322         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10323         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10324         support.
10325         * config/mmix/mmix.h: Likewise.
10326         * config/nds32/nds32.c: Likewise.
10327         * config/stormy/storym16.h: Likewise.
10328         * config/visium/visium.h: Likewise.
10329         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10331 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10333         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10334         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10335         FRAME_POINTER_REGNUM point at high end of local var area.
10337 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10339         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10340         Move comment around.  Do not reset best_edge for a copiable
10341         destination if the copy would cause a partition change.
10342         (better_edge_p): Remove redundant check.
10344 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10346         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10348 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10350         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10351         for math functions that have _Float<N> and _Float<N>X variants.
10352         (mathfn_built_in_2): Add support for math functions that have
10353         _Float<N> and _Float<N>X variants.
10354         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10355         (expand_builtin_mathfn_ternary): Add support for fma with
10356         _Float<N> and _Float<N>X variants.
10357         (expand_builtin): Likewise.
10358         (fold_builtin_3): Likewise.
10359         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10360         create math function _Float<N> and _Float<N>X variants as external
10361         library builtins.
10362         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10363         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10364         the __builtin_ prefix and if not strict ansi, without the prefix.
10365         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10366         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10367         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10368         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10369         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10370         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10371         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10372         function signatures for fma _Float<N> and _Float<N>X variants.
10373         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10374         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10375         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10376         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10377         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10378         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10379         * gencfn-macros.c (print_case_cfn): Add support for math functions
10380         that have _Float<N> and _Float<N>X variants.
10381         (print_define_operator_list): Likewise.
10382         (fltfn_suffixes): Likewise.
10383         (main): Likewise.
10384         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10385         for math functions that have _Float<N> and _Float<N>X variants.
10386         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10387         and _Float<N>X variants.
10388         (COPYSIGN): Likewise.
10389         (FMIN): Likewise.
10390         (FMAX): Likewise.
10391         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10392         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10393         variants.
10394         (integer_valued_read_call_p): Likewise.
10395         * fold-const-call.c (fold_const_call_ss): Likewise.
10396         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10397         _Float<N> and _Float<N>X variants.
10398         (fold_const_call_ssss): Add support for fma _Float<N> and
10399         _Float<N>X variants.
10400         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10401         support for copysign and fma _Float<N> and _Float<N>X variants.
10402         (backprop::process_builtin_call_use): Likewise.
10403         * tree-call-cdce.c (can_test_argument_range); Add support for
10404         sqrt _Float<N> and _Float<N>X variants.
10405         (edom_only_function): Likewise.
10406         (get_no_error_domain): Likewise.
10407         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10408         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10409         copysign _Float<N> and _Float<N>X variants.
10410         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10411         handled by machine independent code.
10412         (FMAF128): Likewise.
10413         * doc/cpp.texi (Common Predefined Macros): Document defining
10414         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10415         fma _Float<N> and _Float<N>X variants.
10417 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10419         PR target/82692
10420         * config/i386/i386-modes.def (CCFPU): Remove definition.
10421         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10422         (ix86_cc_modes_compatible): Ditto.
10423         (ix86_expand_carry_flag_compare): Ditto.
10424         (ix86_expand_int_movcc): Ditto.
10425         (ix86_expand_int_addcc): Ditto.
10426         (ix86_reverse_condition): Ditto.
10427         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10428         Return true/false for unordered/ordered fp comparisons.
10429         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10430         (ix86_prepare_fp_compare_args): Update for rename.
10431         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10432         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10433         (ix86_expand_sse_compare_and_jump): Ditto.
10434         * config/i386/predicates.md (fcmov_comparison_operator):
10435         Remove CCFPU mode handling.
10436         (ix86_comparison_operator): Ditto.
10437         (ix86_carry_flag_operator): Ditto.
10438         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10439         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10440         (*cmpu<mode>_cc_i387): Ditto.
10441         (FPCMP): Remove mode iterator.
10442         (unord): Remove mode attribute.
10443         (unord_subst): New define_subst transformation
10444         (unord): New define_subst attribute.
10445         (unordered): Ditto.
10446         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10447         (*cmpi<unord>xf_i387): Ditto.
10448         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10449         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10450         using unord_subst transformation.
10451         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10452         (round_saeonly): Also handle CCFP mode.
10453         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10454         Remove UNSPEC_SAHF unspec handling.
10456 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10458         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10460 2017-10-27  Jeff Law  <law@redhat.com>
10462         * gimple-ssa-sprintf.c: Include domwalk.h.
10463         (class sprintf_dom_walker): New class, derived from dom_walker.
10464         (sprintf_dom_walker::before_dom_children): New function.
10465         (struct call_info): Moved into sprintf_dom_walker class
10466         (compute_formath_length, handle_gimple_call): Likewise.
10467         (sprintf_length::execute): Call the dominator walker rather
10468         than walking the statements.
10470         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10471         gimple statement locations.
10472         (check_array_bounds): Corresponding changes.  Get the statement's
10473         location directly from wi->stmt.
10475 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10477         PR target/82717
10478         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10480 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10482         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10483         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10485 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10487         PR target/82703
10488         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10489         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10490         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10491         maybe_get_pool_constant.
10492         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10493         Likewise.
10495 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10497         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10498         2.26 caveat.  Update gas and gld versions.
10499         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10500         reference.
10502 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10504         * cgraph.h (set_malloc_flag): Declare.
10505         * cgraph.c (set_malloc_flag_1): New function.
10506         (set_malloc_flag): Likewise.
10507         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10508         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10509         false.
10510         (read_ipa_call_summary): Add support for reading is_return_callee.
10511         (write_ipa_call_summary): Stream is_return_callee.
10512         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10513         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10514         ipa-prop.h, ipa-fnsummary.h.
10515         (pure_const_names): Change to static.
10516         (malloc_state_e): Define.
10517         (malloc_state_names): Define.
10518         (funct_state_d): Add field malloc_state.
10519         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10520         (check_retval_uses): New function.
10521         (malloc_candidate_p): Likewise.
10522         (analyze_function): Add support for malloc attribute.
10523         (pure_const_write_summary): Stream malloc_state.
10524         (pure_const_read_summary): Add support for reading malloc_state.
10525         (dump_malloc_lattice): New function.
10526         (propagate_malloc): New function.
10527         (warn_function_malloc): New function.
10528         (ipa_pure_const::execute): Call propagate_malloc and
10529         ipa_free_fn_summary.
10530         (pass_local_pure_const::execute): Add support for malloc attribute.
10531         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10532         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10534 2017-10-27  Martin Liska  <mliska@suse.cz>
10536         PR gcov-profile/82457
10537         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10538         for fork-like functions to be properly instrumented.
10540 2017-10-27  Richard Biener  <rguenther@suse.de>
10542         PR middle-end/81659
10543         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10544         info when we redirected EH.
10546 2017-10-26  Michael Collison  <michael.collison@arm.com>
10548         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10549         New pattern.
10550         (<optab>_trunchf<GPI:mode>2: New pattern.
10551         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10552         * config/aarch64/iterators.md (wv): New mode attribute.
10553         (vf, VF): New mode attributes.
10554         (vgp, VGP): New mode attributes.
10555         (s): Update attribute with SImode and DImode prefixes.
10557 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10559         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10560         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10561         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10562         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10563         addressing with PIC.
10564         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10565         (nios2_symbolic_constant_p): Likewise.
10566         (nios2_legitimate_address_p): Likewise.
10567         (nios2_r0rel_section_name_p): New.
10568         (nios2_symbol_ref_in_r0rel_data_p): New.
10569         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10570         (r0rel_constant_p): New.
10571         (nios2_print_operand_address): Handle r0rel_constant_p.
10572         (nios2_cdx_narrow_form_p): Likewise.
10573         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10574         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10575         (Nios II Options): Document -mr0rel-sec.
10577 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10579         * config/nios2/nios2.c: Include xregex.h.
10580         (nios2_gprel_sec_regex): New.
10581         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10582         addressing with PIC.
10583         (nios2_small_section_name_p): Check for regex match.
10584         * config/nios2/nios2.opt (mgprel-sec=): New option.
10585         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10586         (Nios II Options): Document -mgprel-sec.
10588 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10590         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10592 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10594         PR tree-optimization/82707
10595         * gimple.c (gimple_copy): Fix unsharing of
10596         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10598 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10600         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10601         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10602         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10603         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10604         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10605         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10606         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10607         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10608         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10609         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10610         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10611         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10612         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10613         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10614         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10615         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10616         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10618 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10620         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10621         default to IBM.
10622         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10623         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10624         warning to rs6000.c.  Remove the Undocumented flag, since it has
10625         been documented.
10626         (-mabi=ibmlongdouble): Likewise.
10627         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10628         already set, set the default format for long double.
10629         (rs6000_debug_reg_global): Print whether long double is IBM or
10630         IEEE.
10631         (rs6000_option_override_internal): Rework setting long double
10632         format.  Only warn if the user is changing the long double default
10633         and they did not use -Wno-psabi.
10634         * doc/invoke.texi (PowerPC options): Update the documentation for
10635         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10637 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10638             Alan Hayward  <alan.hayward@arm.com>
10639             David Sherwood  <david.sherwood@arm.com>
10641         * rtl.h (wider_subreg_mode): New function.
10642         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10643         rather than an unsigned int *.
10644         * ira-color.c (regno_max_ref_width): Replace with...
10645         (regno_max_ref_mode): ...this new variable.
10646         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10647         Use wider_subreg_mode.
10648         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10649         rather than an unsigned int *.
10650         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10651         (process_alt_operands): Likewise.
10652         (invariant_p): Likewise.
10653         * lra-spills.c (assign_mem_slot): Likewise.
10654         (add_pseudo_to_slot): Likewise.
10655         * lra.c (collect_non_operand_hard_regs): Likewise.
10656         (add_regs_to_insn_regno_info): Likewise.
10657         * reload1.c (regno_max_ref_width): Replace with...
10658         (regno_max_ref_mode): ...this new variable.
10659         (reload): Update accordingly.  Update call to
10660         ira_sort_regnos_for_alter_reg.
10661         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10662         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10663         (scan_paradoxical_subregs): Likewise.
10665 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10667         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10668         (aarch64_frame): Add emit_frame_chain boolean.
10669         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10670         Move eh_return case to aarch64_layout_frame.
10671         (aarch64_layout_frame): Initialize emit_frame_chain.
10672         (aarch64_expand_prologue): Use emit_frame_chain.
10674 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10676         * config/aarch64/aarch64.c (aarch64_layout_frame):
10677         Ensure LR is always stored at the bottom of the callee-saves.
10678         Remove rarely used frame layout which saves callee-saves at top of
10679         frame, so the store of LR can be used as a valid probe in all cases.
10681 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10683         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10684         Improve unaligned TImode/TFmode base/offset split.
10686 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10687             Alan Hayward  <alan.hayward@arm.com>
10688             David Sherwood  <david.sherwood@arm.com>
10690         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10691         * combine.c (find_single_use_1): Likewise.
10692         (expand_field_assignment): Likewise.
10693         (move_deaths): Likewise.
10694         * lra-constraints.c (simplify_operand_subreg): Likewise.
10695         (curr_insn_transform): Likewise.
10696         * lra.c (collect_non_operand_hard_regs): Likewise.
10697         (add_regs_to_insn_regno_info): Likewise.
10698         * rtlanal.c (reg_referenced_p): Likewise.
10699         (covers_regno_no_parallel_p): Likewise.
10701 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10703         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10704         Don't print any bits outside the precision of the value.
10705         * wide-int.cc (test_printing): Add some new tests.
10707 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10709         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10710         supports -xbrace_comment option.
10711         * configure: Regenerate.
10712         * config.in: Regenerate.
10713         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10714         (ASM_CPU_SPEC): Use it.
10716 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10718         * target.def (static_rtx_alignment): New hook.
10719         * targhooks.h (default_static_rtx_alignment): Declare.
10720         * targhooks.c (default_static_rtx_alignment): New function.
10721         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10722         * doc/tm.texi: Regenerate.
10723         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10724         instead of targetm.constant_alignment.  Remove call to
10725         set_mem_attributes.
10726         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10727         (cris_preferred_mininum_alignment): New function, split out from...
10728         (cris_constant_alignment): ...here.
10729         (cris_static_rtx_alignment): New function.
10730         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10731         split out from...
10732         (ix86_constant_alignment): ...here.
10733         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10734         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10735         (mmix_static_rtx_alignment): New function.
10736         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10737         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10739 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10741         PR target/81800
10742         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10743         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10745 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10747         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10748         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10749         mark as a sign-extending load.
10750         (local_pic_load_u): Define.
10752 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10754         PR middle-end/82062
10755         * fold-const.c (operand_equal_for_comparison_p): Also return true
10756         if ARG0 is a simple variant of ARG1 with narrower precision.
10757         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10759 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10761         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10762         cost correctly.
10763         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10764         scatter_static, scatter_per_elt.
10765         * x86-tune-costs.h: Add new cost entries.
10767 2017-10-25  Richard Biener  <rguenther@suse.de>
10769         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10770         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10771         class pass_fre): Move to ...
10772         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10773         class pass_fre): ... here and adjust for statistics.
10775 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10777         PR libstdc++/81706
10778         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10779         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10780         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10781         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10782         declarations.
10784 2017-10-25  Richard Biener  <rguenther@suse.de>
10786         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10787         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10788         eliminate_push_avail, eliminate_insert): Move inside...
10789         (class eliminate_dom_walker): ... this class in preparation
10790         of move.
10791         (fini_eliminate): Remove by merging with ...
10792         (eliminate): ... this function.  Adjust for class changes.
10793         (pass_pre::execute): Remove fini_eliminate call.
10794         (pass_fre::execute): Likewise.
10796 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10798         PR target/82460
10799         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10800         (VPERMI2, VPERMI2I): New mode iterators.
10801         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10802         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10803         patterns.
10804         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10805         mode iterator.  Remove 3 old define_insn patterns.
10806         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10807         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10808         VPERMI2 mode iterator, remove the other two expanders.
10809         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10810         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10811         instructions, remove the other two patterns.
10812         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10813         mode iterator, remove the other two patterns.
10814         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10815         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10816         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10817         and adjust argument order accordingly.
10818         (ix86_expand_vec_perm): Adjust caller.
10819         (expand_vec_perm_1): Likewise.
10820         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10821         (expand_vec_perm_vpermt2_vpshub2): ... this.
10822         (ix86_expand_vec_perm_const_1): Adjust caller.
10823         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10825         PR target/82370
10826         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10827         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10828         (vec_shl_<mode>): Remove unused expander.
10829         (avx512bw_<shift_insn><mode>3): New define_insn.
10830         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10831         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10833 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10835         PR c++/82466
10836         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10837         description.
10839 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10841         PR rtl-optimization/82396
10842         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10843         (autopref_multipass_init): Simplify initialization.
10844         (autopref_rank_data): Simplify sort order.
10845         * gcc/sched-int.h (autopref_multipass_data_): Remove
10846         multi_mem_insn_p, min_offset and max_offset.
10848 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10850         PR middle-end/60580
10851         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10852         Check special value of flag_omit_frame_pointer.
10853         (aarch64_can_eliminate): Likewise.
10854         (aarch64_override_options_after_change_1): Simplify handling of
10855         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10857 2017-10-24  Richard Biener  <rguenther@suse.de>
10859         PR tree-optimization/82697
10860         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10861         zero for conditional load and unconditional store.
10863 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10865         * doc/install.texi: Document bootstrap-cet.
10867 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10869         PR target/82659
10870         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
10871         ENDBR instruction at function entrance if function is only
10872         called directly.
10874 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10876         PR target/82628
10877         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
10878         patterns to better describe from which operation the CF is computed.
10879         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
10880         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
10881         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
10882         is 0, use _0 suffixed expanders instead of emitting a comparison
10883         before it.
10885 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
10887         * config/i386/i386.md(*movsf_internal, *movdf_internal):
10888         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
10890 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
10892         PR middle-end/82569
10893         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
10894         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
10895         * loop-iv.c (iv_get_reaching_def): Likewise.
10896         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
10897         variable is promoted and the partition contains undefined values.
10899 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10901         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
10902         reflect reality.
10903         (nios2_address_cost): Define.
10904         (nios2_legitimize_address): Recognize (exp + constant) directly.
10905         (TARGET_ADDRESS_COST): Define.
10907 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10909         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
10910         (nios2_symbolic_memory_operand_p): Declare.
10911         (nios2_split_large_constant): Declare.
10912         (nios2_split_symbolic_memory_operand): Declare.
10913         * config/nios2/nios2.c: Adjust includes.
10914         (nios2_symbolic_constant_allowed): New.
10915         (nios2_symbolic_constant_p): New.
10916         (nios2_plus_symbolic_constant_p): New.
10917         (nios2_valid_addr_expr_p): Recognize addresses involving
10918         symbolic constants.
10919         (nios2_legitimate_address_p): Likewise, also LO_SUM.
10920         (nios2_symbolic_memory_operand_p): New.
10921         (nios2_large_constant_p): New.
10922         (nios2_split_large_constant): New.
10923         (nios2_split_plus_large_constant): New.
10924         (nios2_split_symbolic_memory_operand): New.
10925         (nios2_legitimize_address): Code refactoring.  Handle addresses
10926         involving symbolic constants.
10927         (nios2_emit_move_sequence): Likewise.
10928         (nios2_print_operand): Improve error output.
10929         (nios2_print_operand_address): Handle LO_SUM.
10930         (nios2_cdx_narrow_form_p): Likewise.
10931         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
10932         operands involving symbolic constants.
10933         (movhi_internal, movsi_internal): Likewise.
10934         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
10935         (extendhisi2, extendqi<mode>2): Likewise.
10937 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10939         * tree-pass.h (PROP_rtl_split_insns): Define.
10940         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
10942 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10944         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
10946 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
10948         PR debug/82630
10949         * target.def (const_not_ok_for_debug_p): Default to
10950         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
10951         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
10952         * targhooks.c (default_const_not_ok_for_debug_p): New function.
10953         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
10954         which targetm.const_not_ok_for_debug_p returned true.
10955         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
10956         for UNSPECs.
10957         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
10958         Likewise.
10959         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
10960         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
10961         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
10962         if !base_term_p.
10963         (ix86_const_not_ok_for_debug_p): New function.
10964         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
10965         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
10967 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
10969         PR bootstrap/82610
10970         * system.h: Conditionally include "unique-ptr.h" if
10971         INCLUDE_UNIQUE_PTR is defined.
10972         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
10973         of defining INCLUDE_UNIQUE_PTR before including "system.h".
10975 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
10977         * config/rl78/rl78.md: New define_expand "subdi3".
10979 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
10981         PR target/82673
10982         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
10983         DF_REF_INSN if DF_REF_INSN_INFO is false.
10985 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
10987         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
10988         xmm_move instead of sse_move.
10989         (sse_store_index): New function.
10990         (ix86_register_move_cost): Be more sensible about mismatch stall;
10991         model AVX moves correctly; make difference between sse->integer and
10992         integer->sse.
10993         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
10994         moves; make difference between SSE and AVX.
10995         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
10996         and zmm_move. Increase size of sse load and store tables;
10997         add unaligned load and store tables; add ssemmx_to_integer.
10998         * x86-tune-costs.h: Update all entries according to real
10999         move latencies from Agner Fog's manual and chip documentation.
11001 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11003         PR target/82628
11004         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11005         * config/i386/constraints.md (Wf): New constraint.
11006         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11007         (cmp<dwi>_doubleword): Removed.
11008         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11009         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11010         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11011         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11012         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11014         * common.opt (gcolumn-info): Enable by default.
11015         * doc/invoke.texi (gcolumn-info): Document new default.
11017 2017-10-23  Richard Biener  <rguenther@suse.de>
11019         PR tree-optimization/82672
11020         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11021         Fold the stmt if we propagated into it.
11023 2017-10-23  Richard Biener  <rguenther@suse.de>
11025         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11026         (bitmap_remove_expr_from_set): ... this.  All callers call this
11027         for non-constant values.
11028         (bitmap_set_subtract): Rename to...
11029         (bitmap_set_subtract_expressions): ... this.  Adjust and
11030         optimize.
11031         (bitmap_set_contains_value): Remove superfluous check.
11032         (bitmap_set_replace_value): Inline into single caller ...
11033         (bitmap_value_replace_in_set): ... here and simplify.
11034         (dependent_clean): Merge into ...
11035         (clean): ... this using an overload.  Adjust.
11036         (prune_clobbered_mems): Adjust.
11037         (compute_antic_aux): Likewise.
11038         (compute_partial_antic_aux): Likewise.
11040 2017-10-23  Richard Biener  <rguenther@suse.de>
11042         PR tree-optimization/82129
11043         Revert
11044         2017-08-01  Richard Biener  <rguenther@suse.de>
11046         PR tree-optimization/81181
11047         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11048         (compute_antic): ... end of iteration here.
11050 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11052         * target.def (starting_frame_offset): New hook.
11053         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11054         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11055         * doc/tm.texi.in: Regenerate.
11056         * hooks.h (hook_hwi_void_0): Declare.
11057         * hooks.c (hook_hwi_void_0): New function.
11058         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11059         STARTING_FRAME_OFFSET.
11060         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11061         * reload1.c (reload): Likewise.
11062         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11063         instead of STARTING_FRAME_OFFSET.
11064         * function.c (try_fit_stack_local): Likewise.
11065         (assign_stack_local_1): Likewise
11066         (instantiate_virtual_regs): Likewise.
11067         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11068         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11069         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11070         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11071         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11072         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11073         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11074         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11075         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11076         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11077         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11078         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11079         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11080         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11081         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11082         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11083         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11084         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11085         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11086         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11087         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11088         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11089         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11090         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11091         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11092         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11093         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11094         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11095         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11096         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11097         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11098         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11099         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11100         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11101         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11102         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11103         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11104         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11105         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11106         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11107         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11108         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11109         return a HOST_WIDE_INT.
11110         (avr_builtin_setjmp_frame_value): Use it instead of
11111         STARTING_FRAME_OFFSET.
11112         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11113         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11114         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11115         New function.
11116         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11117         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11118         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11119         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11120         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11121         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11122         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11123         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11124         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11125         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11126         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11127         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11128         New function.
11129         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11130         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11131         * config/mips/mips.c (mips_compute_frame_info): Refer to
11132         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11133         (mips_starting_frame_offset): New function.
11134         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11135         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11136         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11137         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11138         and return a HOST_WIDE_INT.
11139         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11140         (mmix_initial_elimination_offset): Refer to
11141         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11142         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11143         * config/pa/pa.c (pa_starting_frame_offset): New function.
11144         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11145         (pa_expand_prologue): Likewise.
11146         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11147         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11148         !FRAME_GROWS_DOWNWARD handling to...
11149         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11150         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11151         !FRAME_GROWS_DOWNWARD handling to...
11152         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11153         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11154         !FRAME_GROWS_DOWNWARD handling to...
11155         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11156         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11157         Redefine.
11158         (rs6000_starting_frame_offset): New function.
11159         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11160         !FRAME_GROWS_DOWNWARD handling to...
11161         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11162         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11163         !FRAME_GROWS_DOWNWARD handling to...
11164         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11165         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11166         !FRAME_GROWS_DOWNWARD handling to...
11167         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11168         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11169         (rs6000_starting_frame_offset): New function.
11170         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11171         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11172         * config/vax/vax.c (vax_starting_frame_offset): New function.
11173         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11174         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11175         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11176         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11177         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11178         * system.h (STARTING_FRAME_OFFSET): Poison.
11180 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11182         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11183         SCALAR_TYPE_MODE instead of TYPE_MODE.
11185 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11186             Alan Hayward  <alan.hayward@arm.com>
11187             David Sherwood  <david.sherwood@arm.com>
11189         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11191 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11192             Alan Hayward  <alan.hayward@arm.com>
11193             David Sherwood  <david.sherwood@arm.com>
11195         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11197 2017-10-23  Richard Biener  <rguenther@suse.de>
11199         PR tree-optimization/82129
11200         * tree-ssa-pre.c (bitmap_set_and): Remove.
11201         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11202         canonicalizing expressions in the set to those with lowest
11203         ID rather than taking that from the first edge.
11205 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11207         * combine.c (rtx_equal_for_field_assignment_p): Use
11208         byte_lowpart_offset.
11210 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11211             Alan Hayward  <alan.hayward@arm.com>
11212             David Sherwood  <david.sherwood@arm.com>
11214         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11215         to a SUBREG_PROMOTED_VAR.
11217 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11218             Alan Hayward  <alan.hayward@arm.com>
11219             David Sherwood  <david.sherwood@arm.com>
11221         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11222         (expand_debug_source_expr): Likewise.
11223         * combine.c (combine_simplify_rtx): Likewise.
11224         * cse.c (fold_rtx): Likewise.
11225         * optabs.c (expand_float): Likewise.
11226         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11227         (simplify_binary_operation_1): Likewise.
11229 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11230             Alan Hayward  <alan.hayward@arm.com>
11231             David Sherwood  <david.sherwood@arm.com>
11233         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11234         (record_promoted_value): Likewise.
11235         * expr.c (expand_expr_real_2): Likewise.
11236         * ree.c (update_reg_equal_equiv_notes): Likewise.
11237         (combine_set_extension): Likewise.
11238         * rtlanal.c (low_bitmask_len): Likewise.
11239         * simplify-rtx.c (neg_const_int): Likewise.
11240         (simplify_binary_operation_1): Likewise.
11242 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11243             Alan Hayward  <alan.hayward@arm.com>
11244             David Sherwood  <david.sherwood@arm.com>
11246         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11247         * regcprop.c (maybe_mode_change): Likewise.
11248         * reload1.c (alter_reg): Likewise.
11250 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11252         * inchash.h (inchash::hash::add_wide_int): New function.
11253         * lto-streamer-out.c (hash_tree): Use it.
11255 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11257         * inchash.h (inchash::hash::add_wide_int): Rename to...
11258         (inchash::hash::add_hwi): ...this.
11259         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11260         (polymorphic_call_target_hasher::hash): Likewise.
11261         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11262         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11263         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11264         * lto-streamer-out.c (hash_tree): Likewise.
11265         * optc-save-gen.awk: Likewise.
11266         * tree.c (add_expr): Likewise.
11268 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11270         PR target/52451
11271         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11272         for ordered inequality comparisons even with TARGET_IEEE_FP.
11274 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11276         PR target/82628
11277         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11278         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11279         Expand with cmp<dwi>_doubleword.
11281 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11283         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11284         List CET intrinsics.
11285         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11286         specific to -fcf-protection option.
11288 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11290         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11291         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11292         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11293         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11294         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11295         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11296         (extra_objs): Add cet.o for Linux/x86 targets.
11297         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11298         * config/i386/cet.c: New file.
11299         * config/i386/cetintrin.h: Likewise.
11300         * config/i386/t-cet: Likewise.
11301         * config/i386/cpuid.h (bit_SHSTK): New.
11302         (bit_IBT): Likewise.
11303         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11304         pass IBT and SHSTK bits.
11305         * config/i386/i386-builtin-types.def
11306         (VOID_FTYPE_UNSIGNED_PVOID): New.
11307         (VOID_FTYPE_UINT64_PVOID): Likewise.
11308         * config/i386/i386-builtin.def: Add CET intrinsics.
11309         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11310         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11311         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11312         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11313         prototype.
11314         * config/i386/i386.c (rest_of_insert_endbranch): New.
11315         (pass_data_insert_endbranch): Likewise.
11316         (pass_insert_endbranch): Likewise.
11317         (make_pass_insert_endbranch): Likewise.
11318         (ix86_notrack_prefixed_insn_p): Likewise.
11319         (ix86_target_string): Add -mibt, -mshstk flags.
11320         (ix86_option_override_internal): Add flag_cf_protection
11321         processing.
11322         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11323         (ix86_print_operand): Add 'notrack' prefix output.
11324         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11325         (ix86_expand_builtin): Expand CET intrinsics.
11326         (x86_output_mi_thunk): Add 'endbranch' instruction.
11327         * config/i386/i386.h (TARGET_IBT): New.
11328         (TARGET_IBT_P): Likewise.
11329         (TARGET_SHSTK): Likewise.
11330         (TARGET_SHSTK_P): Likewise.
11331         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11332         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11333         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11334         (builtin_setjmp_setup): New pattern.
11335         (builtin_longjmp): Likewise.
11336         (rdssp<mode>): Likewise.
11337         (incssp<mode>): Likewise.
11338         (saveprevssp): Likewise.
11339         (rstorssp): Likewise.
11340         (wrss<mode>): Likewise.
11341         (wruss<mode>): Likewise.
11342         (setssbsy): Likewise.
11343         (clrssbsy): Likewise.
11344         (nop_endbr): Likewise.
11345         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11346         options.
11347         * config/i386/immintrin.h: Include <cetintrin.h>.
11348         * config/i386/linux-common.h
11349         (file_end_indicate_exec_stack_and_cet): New prototype.
11350         (TARGET_ASM_FILE_END): New.
11352 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11354         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11355         latencies instead of having separate table; make difference between
11356         integer and float costs.
11357         * i386.h (processor_costs): Remove scalar_stmt_cost,
11358         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11359         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11360         vec_store_cost.
11361         * x86-tune-costs.h: Remove entries which has been removed in
11362         procesor_costs from all tables; make cond_taken_branch_cost
11363         and cond_not_taken_branch_cost COST_N_INSNS based.
11365 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11367         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11369 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11371         * config/i386/i386.md (isa): Remove fma_avx512f.
11372         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11373         <avx512>_fmadd_<mode>_mask3<round_name>,
11374         <avx512>_fmsub_<mode>_mask<round_name>,
11375         <avx512>_fmsub_<mode>_mask3<round_name>,
11376         <avx512>_fnmadd_<mode>_mask<round_name>,
11377         <avx512>_fnmadd_<mode>_mask3<round_name>,
11378         <avx512>_fnmsub_<mode>_mask<round_name>,
11379         <avx512>_fnmsub_<mode>_mask3<round_name>,
11380         <avx512>_fmaddsub_<mode>_mask<round_name>,
11381         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11382         <avx512>_fmsubadd_<mode>_mask<round_name>,
11383         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11384         (*vec_widen_umult_even_v16si<mask_name>,
11385         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11386         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11388 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11390         * extend.texi: Add 'nocf_check' documentation.
11391         * gimple.texi: Add second parameter to
11392         gimple_build_call_from_tree.
11393         * invoke.texi: Add -fcf-protection documentation.
11394         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11396 2017-10-20  Richard Biener  <rguenther@suse.de>
11398         PR tree-optimization/82473
11399         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11400         the largest input type.
11402 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11404         * c-attribs.c (handle_nocf_check_attribute): New function.
11405         (c_common_attribute_table): Add 'nocf_check' handling.
11406         * gimple-parser.c: Add second argument NULL to
11407         gimple_build_call_from_tree.
11408         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11409         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11410         call insn.
11411         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11412         * common.opt: Add fcf-protection flag.
11413         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11414         * flag-types.h: Add enum cf_protection_level.
11415         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11416         Add 'nocf_check' attribute propagation to gimple call.
11417         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11418         (gimple_build_call_from_tree): Update prototype.
11419         (gimple_call_nocf_check_p): New function.
11420         (gimple_call_set_nocf_check): Likewise.
11421         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11422         * ipa-icf.c: Add nocf_check attribute in statement hash.
11423         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11424         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11425         * toplev.c (process_options): Add flag_cf_protection handling.
11427 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11429         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11431 2017-10-20  Richard Biener  <rguenther@suse.de>
11433         PR tree-optimization/82603
11434         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11435         remove false predicated stores.
11437 2017-10-20  Richard Biener  <rguenther@suse.de>
11439         * graphite-isl-ast-to-gimple.c
11440         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11441         Remove return value and simplify, dump copied stmt after lhs
11442         adjustment.
11443         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11444         Reduce dump verbosity.
11445         (gsi_insert_earliest): Likewise.
11446         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11447         * graphite.c (print_global_statistics): Adjust dumping.
11448         (print_graphite_scop_statistics): Likewise.
11449         (print_graphite_statistics): Do not dump loops here.
11450         (graphite_transform_loops): But here.
11452 2017-10-20  Nicolas Roche  <roche@adacore.com>
11454         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11455         * configure: Regenerate.
11457 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11459         PR target/82158
11460         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11461         functions when optimizing replace GIMPLE_RETURN stmts with
11462         calls to __builtin_unreachable ().
11464         PR sanitizer/82595
11465         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11466         for -fsanitize=thread link of executables.
11467         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11468         link of executables.
11470         PR target/82370
11471         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11472         New mode iterators.
11473         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11474         define_insns for logical vector shifts to use VI248_AVX512BW
11475         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11476         condition, useless isa and prefix attributes.  Change the first
11477         2 of these define_insns to ...
11478         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11479         define_insn for avx512vl.
11480         (<shift_insn><mode>3): ... and this, new define_insn without
11481         masking for non-avx512vl.
11483         PR target/82370
11484         * config/i386/sse.md (*andnot<mode>3,
11485         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11486         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11487         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11488         not applied use empty suffix even for TARGET_AVX512VL.
11489         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11490         is applied, supply evex,evex or evex,evex,evex instead of just
11491         evex.
11493 2017-10-20  Julia Koval  <julia.koval@intel.com>
11495         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11496         (OPTION_MASK_ISA_GFNI_UNSET): New.
11497         (ix86_handle_option): Handle OPT_mgfni.
11498         * config/i386/cpuid.h (bit_GFNI): New.
11499         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11500         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11501         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11502         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11503         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11504         * config/i386/i386.opt: Add mgfni.
11506 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11508         * config/msp430/msp430.c (msp430_option_override): Disable
11509         -fdelete-null-pointer-checks.
11510         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11512 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11514         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11515         of x87 and SSE instructions.
11517 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11519         * asan.c (create_cond_insert_point): Do not update edge count.
11520         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11521         (afdo_propagate_circuit): Likewise.
11522         (afdo_calculate_branch_prob): Likewise.
11523         (afdo_annotate_cfg): Likewise.
11524         * basic-block.h (struct edge_def): Remove count.
11525         (edge_def::count): New accessor.
11526         * bb-reorder.c (rotate_loop): Update.
11527         (find_traces_1_round): Update.
11528         (connect_traces): Update.
11529         (sanitize_hot_paths): Update.
11530         * cfg.c (unchecked_make_edge): Update.
11531         (make_single_succ_edge): Update.
11532         (check_bb_profile): Update.
11533         (dump_edge_info): Update.
11534         (update_bb_profile_for_threading): Update.
11535         (scale_bbs_frequencies_int): Update.
11536         (scale_bbs_frequencies_gcov_type): Update.
11537         (scale_bbs_frequencies_profile_count): Update.
11538         (scale_bbs_frequencies): Update.
11539         * cfganal.c (connect_infinite_loops_to_exit): Update.
11540         * cfgbuild.c (compute_outgoing_frequencies): Update.
11541         (find_many_sub_basic_blocks): Update.
11542         * cfgcleanup.c (try_forward_edges): Update.
11543         (try_crossjump_to_edge): Update
11544         * cfgexpand.c (expand_gimple_cond): Update
11545         (expand_gimple_tailcall): Update
11546         (construct_exit_block): Update
11547         * cfghooks.c (verify_flow_info): Update
11548         (redirect_edge_succ_nodup): Update
11549         (split_edge): Update
11550         (make_forwarder_block): Update
11551         (duplicate_block): Update
11552         (account_profile_record): Update
11553         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11554         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11555         * cfgloopmanip.c (scale_loop_profile): Update.
11556         (loopify): Update.
11557         (lv_adjust_loop_entry_edge): Update.
11558         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11559         (force_nonfallthru_and_redirect): Update.
11560         (purge_dead_edges): Update.
11561         (rtl_flow_call_edges_add): Update.
11562         * cgraphunit.c (init_lowered_empty_function): Update.
11563         (cgraph_node::expand_thunk): Update.
11564         * gimple-pretty-print.c (dump_probability): Update.
11565         (dump_edge_probability): Update.
11566         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11567         * haifa-sched.c (sched_create_recovery_edges): Update.
11568         * hsa-gen.c (convert_switch_statements): Update.
11569         * ifcvt.c (dead_or_predicable): Update.
11570         * ipa-inline-transform.c (inline_transform): Update.
11571         * ipa-split.c (split_function): Update.
11572         * ipa-utils.c (ipa_merge_profiles): Update.
11573         * loop-doloop.c (add_test): Update.
11574         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11575         * lto-streamer-in.c (input_cfg): Update.
11576         (input_function): Update.
11577         * lto-streamer-out.c (output_cfg): Update.
11578         * modulo-sched.c (sms_schedule): Update.
11579         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11580         * predict.c (maybe_hot_edge_p): Update.
11581         (unlikely_executed_edge_p): Update.
11582         (probably_never_executed_edge_p): Update.
11583         (dump_prediction): Update.
11584         (drop_profile): Update.
11585         (propagate_unlikely_bbs_forward): Update.
11586         (determine_unlikely_bbs): Update.
11587         (force_edge_cold): Update.
11588         * profile.c (compute_branch_probabilities): Update.
11589         * reg-stack.c (better_edge): Update.
11590         * shrink-wrap.c (handle_simple_exit): Update.
11591         * tracer.c (better_p): Update.
11592         * trans-mem.c (expand_transaction): Update.
11593         (split_bb_make_tm_edge): Update.
11594         * tree-call-cdce.c: Update.
11595         * tree-cfg.c (gimple_find_sub_bbs): Update.
11596         (gimple_split_edge): Update.
11597         (gimple_duplicate_sese_region): Update.
11598         (gimple_duplicate_sese_tail): Update.
11599         (gimple_flow_call_edges_add): Update.
11600         (insert_cond_bb): Update.
11601         (execute_fixup_cfg): Update.
11602         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11603         * tree-complex.c (expand_complex_div_wide): Update.
11604         * tree-eh.c (lower_resx): Update.
11605         (unsplit_eh): Update.
11606         (cleanup_empty_eh_move_lp): Update.
11607         * tree-inline.c (copy_edges_for_bb): Update.
11608         (freqs_to_counts): Update.
11609         (copy_cfg_body): Update.
11610         * tree-ssa-dce.c (remove_dead_stmt): Update.
11611         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11612         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11613         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11614         (unloop_loops): Update.
11615         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11616         * tree-ssa-loop-split.c (connect_loops): Update.
11617         (split_loop): Update.
11618         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11619         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11620         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11621         * tree-ssa-reassoc.c (branch_fixup): Update.
11622         * tree-ssa-tail-merge.c (replace_block_by): Update.
11623         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11624         (compute_path_counts): Update.
11625         (update_profile): Update.
11626         (recompute_probabilities): Update.
11627         (update_joiner_offpath_counts): Update.
11628         (estimated_freqs_path): Update.
11629         (freqs_to_counts_path): Update.
11630         (clear_counts_path): Update.
11631         (ssa_fix_duplicate_block_edges): Update.
11632         (duplicate_thread_path): Update.
11633         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11634         (case_bit_test_cmp): Update.
11635         (collect_switch_conv_info): Update.
11636         (gen_inbound_check): Update.
11637         (do_jump_if_equal): Update.
11638         (emit_cmp_and_jump_insns): Update.
11639         * tree-tailcall.c (decrease_profile): Update.
11640         (eliminate_tail_call): Update.
11641         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11642         (vect_do_peeling): Update.
11643         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11644         * ubsan.c (ubsan_expand_null_ifn): Update.
11645         (ubsan_expand_ptr_ifn): Update.
11646         * value-prof.c (gimple_divmod_fixed_value): Update.
11647         (gimple_mod_pow2): Update.
11648         (gimple_mod_subtract): Update.
11649         (gimple_ic): Update.
11650         (gimple_stringop_fixed_value): Update.
11652 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11654         PR target/82618
11655         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11657 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11659         PR rtl-optimization/82395
11660         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11661         based on non_spilled_static_chain_regno_p.
11663 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11665         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11666         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11667         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11669 2017-10-19  Martin Sebor  <msebor@redhat.com>
11671         PR tree-optimization/82596
11672         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11674 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11676         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11677         (get_mem_refs_of_builtin_call): Likewise.
11678         * builtins.c (expand_builtin_apply): Adjust call to
11679         allocate_dynamic_stack_space.
11680         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11681         the third argument to allocate_dynamic_stack_space, otherwise -1.
11682         (expand_builtin): Deal with all alloca variants.
11683         (is_inexpensive_builtin): Likewise.
11684         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11685         * calls.c (special_function_p): Deal with all alloca variants.
11686         (initialize_argument_information): Adjust call to
11687         allocate_dynamic_stack_space.
11688         (expand_call): Likewise.
11689         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11690         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11691         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11692         use it for the stack usage computation.
11693         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11694         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11695         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11696         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11697         (in_loop_p): Remove first argument and useless check.
11698         (pass_walloca::execute): Remove useless test and adjust call to above.
11699         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11700         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11701         (gimplify_call_expr): Deal with all alloca variants.
11702         * hsa-gen.c (gen_hsa_alloca): Likewise.
11703         (gen_hsa_insns_for_call): Likewise.
11704         * ipa-pure-const.c (special_builtin_state): Likewise.
11705         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11706         * tree-object-size.c (alloc_object_size): Likewise.
11707         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11708         (call_may_clobber_ref_p_1): Likewise.
11709         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11710         (ccp_fold_stmt): Likewise.
11711         (optimize_stack_restore): Likewise.
11712         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11713         (mark_all_reaching_defs_necessary_1): Likewise.
11714         (propagate_necessity): Likewise.
11715         (eliminate_unnecessary_stmts): Likewise.
11716         * tree.c (build_common_builtin_nodes): Build
11717         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11718         (build_alloca_call_expr): New function.
11719         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11720         (CASE_BUILT_IN_ALLOCA): Likewise.
11721         (build_alloca_call_expr): Declare.
11722         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11724 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11726         PR debug/82509
11727         * dwarf2out.c (new_die_raw): New static inline function.
11728         (new_die): Use it to create the DIE.
11729         (add_AT_external_die_ref): Likewise.
11730         (clone_die): Likewise.
11731         (clone_as_declaration): Likewise.
11732         (dwarf2out_vms_debug_main_pointer): Likewise.
11733         (base_type_die): Likewise.  Remove early return for corner cases.
11734         Do not call add_pubtype on the DIE here.
11735         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11736         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11737         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11738         native order exists for base types, attach the DIE manually and call
11739         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11741 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11743         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11744         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11745         the mem into it.
11746         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11748 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11750         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11751         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11752         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11753         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11754         * opts.c (sanitizer_opts): Add builtin.
11755         * ubsan.c (instrument_builtin): New function.
11756         (pass_ubsan::execute): Call it.
11757         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11758         * doc/invoke.texi: Document -fsanitize=builtin.
11760         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11761         builtins, store max (log2 (align), 0) into uchar field instead of
11762         align into uptr field.
11763         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11764         store uchar 0 field instead of uptr 0 field.
11765         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11766         instead of passing one address of struct with 2 locations pass
11767         two addresses of structs with 1 location each.
11768         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11769         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11770         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11771         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11772         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11773         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11774         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11775         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11777 2017-10-19  Martin Liska  <mliska@suse.cz>
11779         PR driver/81829
11780         * file-find.c (remove_prefix): Remove.
11781         * file-find.h (remove_prefix): Likewise.
11782         * gcc-ar.c: Remove smartness of lookup.
11784 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11786         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11787         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11788         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11790 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11792         PR target/82580
11793         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11794         (setcc + and to xor + setcc): New peephole2.
11796 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11798         * doc/sourcebuild.texi (Test Directives, Variants of
11799         dg-require-support): Add dg-require-stack-size.
11801 2017-10-19  Martin Liska  <mliska@suse.cz>
11803         PR sanitizer/82517
11804         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11805         that have a large alignment.
11806         (gimplify_target_expr): Likewise.
11808 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11810         PR rtl-optimization/82602
11811         * ira.c (rtx_moveable_p): Return false for volatile asm.
11813 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11815         PR target/82580
11816         * config/i386/i386-modes.def (CCGZ): New CC mode.
11817         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11818         * config/i386/predicates.md (ix86_comparison_operator):
11819         Handle CCGZmode.
11820         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11821         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11822         with double-word subtraction.
11823         (put_condition_code): Handle CCGZmode.
11825 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11827         * wide-int.cc (debug (const wide_int &)): New.
11828         (debug (const wide_int *)): New.
11829         (debug (const widest_int &)): New.
11830         (debug (const widest_int *)): New.
11832 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11834         PR middle-end/82556
11835         * lra-constraints.c (curr_insn_transform): Use non-input operand
11836         instead of output one for matched reload.
11838 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11840         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11841         (tree-ssa-loop-ivopts.h): New header file.
11842         (struct builtin_info): New fields.
11843         (classify_builtin_1): Compute and record base and offset parts for
11844         memset builtin partition by calling strip_offset.
11845         (offset_cmp, fuse_memset_builtins): New functions.
11846         (finalize_partitions): Fuse adjacent memset partitions by calling
11847         above function.
11848         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11849         Expose the interface.
11850         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11852 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11854         PR tree-optimization/82574
11855         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
11856         that data reference must be executed exactly once per iteration
11857         against the outermost loop in nest.
11858         (classify_partition): Update call to above function.
11860 2017-10-18  Richard Biener  <rguenther@suse.de>
11862         PR tree-optimization/82591
11863         * graphite.c (graphite_transform_loops): Move code gen message
11864         printing ...
11865         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
11866         Here.  Handle scop_to_isl_ast failing.
11867         (scop_to_isl_ast): Limit the number of ISL operations.
11869 2017-10-18  Richard Biener  <rguenther@suse.de>
11871         * graphite-isl-ast-to-gimple.c
11872         (translate_isl_ast_to_gimple::set_rename): Simplify.
11873         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
11874         (graphite_copy_stmts_from_block): ... here.
11875         (copy_bb_and_scalar_dependences): Simplify.
11876         (add_parameters_to_ivs_params): Canonicalize.
11877         (generate_entry_out_of_ssa_copies): Simplify.
11878         * graphite-sese-to-poly.c (extract_affine_name): Simplify
11879         by passing in ISL dimension.
11880         (parameter_index_in_region_1): Rename to ...
11881         (parameter_index_in_region): ... this.
11882         (extract_affine): Adjust assert, pass down parameter index.
11883         (add_param_constraints): Use range-info when available.
11884         (build_scop_context): Adjust.
11885         * sese.c (new_sese_info): Adjust.
11886         (free_sese_info): Likewise.
11887         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
11888         Remove unused typedefs.
11889         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
11891 2017-10-18  Martin Liska  <mliska@suse.cz>
11893         * combine.c (simplify_compare_const): Add gcc_fallthrough.
11895 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11897         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
11898         (s390_sched_init): Do not reset s390_sched_state if we entered the
11899         current basic block via a fallthru edge and all others are unlikely.
11901 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11903         * config/s390/s390.c (NUM_SIDES): New variable.
11904         (LONGRUNNING_THRESHOLD): New variable.
11905         (LATENCY_FACTOR): New variable.
11906         (s390_sched_score): Decrease score for long-running instructions on
11907         wrong side.
11908         (s390_sched_variable_issue): Perform bookkeeping for long-running
11909         instructions.
11911 2017-10-18  Richard Biener  <rguenther@suse.de>
11913         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
11914         Simplify with removal of the parameter rename map.
11915         (set_rename): Likewise.
11916         (should_copy_to_new_region): Likewise.
11917         (graphite_copy_stmts_from_block): Likewise.
11918         (copy_bb_and_scalar_dependences): Remove initialization of
11919         unused copied_bb_map.
11920         (copy_def): Remove.
11921         (copy_internal_parameters): Likewise.
11922         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
11923         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11924         Use INTEGRAL_TYPE_P.
11925         (parameter_index_in_region_1): Rename to ...
11926         (assign_parameter_index_in_region): ... this.  Assert we have
11927         a parameter we handle.
11928         (scan_tree_for_params): Adjust.
11929         * sese.h (parameter_rename_map_t): Remove.
11930         (struct sese_info_t): Remove unused parameter_rename_map and
11931         copied_bb_map members.
11932         * sese.c (new_sese_info): Adjust.
11933         (free_sese_info): Likewise.
11935 2017-10-18  Martin Liska  <mliska@suse.cz>
11937         PR sanitizer/82545
11938         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
11939         on an abnormal edge.
11941 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11943         * doc/invoke.texi (ffunction-sections and fdata-sections):
11944         Update.
11946 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
11948         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
11949         the use statement can throw internally.
11951 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
11953         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
11954         any RTX present on the RHS of a SET.
11955         * compare-elim.c (try_eliminate_compare): Restore comment.
11957 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
11959         * langhooks.h (struct lang_hooks): Document that tree_size langhook
11960         may be also called on tcc_type nodes.
11961         * langhooks.c (lhd_tree_size): Likewise.
11963 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
11965         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
11966         format_warning_at_substring.
11967         (maybe_warn): Convert source_range * param to a location_t.  Pass
11968         UNKNOWN_LOCATION rather than NULL to fmtwarn.
11969         (format_directive): Remove code to extract source_ranges and
11970         source_range * in favor of just a location_t.
11971         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
11972         fmtwarn.
11973         * substring-locations.c (format_warning_va): Convert
11974         source_range * param to a location_t.
11975         (format_warning_at_substring): Likewise.
11976         * substring-locations.h (format_warning_va): Likewise.
11977         (format_warning_at_substring): Likewise.
11979 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
11981         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
11982         vec_scatter_store
11983         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
11984         and scatter/gather ops.
11986         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
11987         vec_gather_load and vec_scatter_store.
11988         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
11989         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
11990         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
11991         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
11992         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
11993         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
11995 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
11997         * reg-stack.c (compare_for_stack_reg): Add bool argument.
11998         Detect FTST instruction and handle its register pops.  Only pop
11999         second operand if can_pop_second_op is true.
12000         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12001         set can_pop_second_op to false in the compare_for_stack_reg call.
12003         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12004         output_fp_compare for stack register operands.
12005         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12006         instructions here.  Do not emit stack register pops here.  Assert
12007         that FCOMPP pops next to top stack register.  Rewrite function.
12009 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12011         PR middle-end/82577
12012         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12013         use DECL_ASSEMBLER_NAME_RAW.
12015         PR middle-end/82546
12016         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12017         TYPE nodes.
12019 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12020             Wilco Dijkstra <wilco.dijkstra@arm.com>
12022         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12023         converstion to Pmode from the buf_addr.
12025 2017-10-17  Richard Biener  <rguenther@suse.de>
12027         * graphite-dependences.c (scop_get_reads_and_writes): Change
12028         output parameters to references.
12030 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12032         PR 71026/tree-optimization
12033         * fold-const.c (distribute_real_division): Removed.
12034         (fold_binary_loc): Remove calls to distribute_real_divison.
12036 2017-10-17  Richard Biener  <rguenther@suse.de>
12038         * graphite-scop-detection.c
12039         (scop_detection::stmt_has_simple_data_refs_p): Always use
12040         the full nest as region.
12041         (try_generate_gimple_bb): Likewise.
12042         * sese.c (scalar_evolution_in_region): Simplify now that
12043         SCEV can handle instantiation in regions.
12044         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12045         in the non-loop part of a function if requested.
12047 2017-10-17  Richard Biener  <rguenther@suse.de>
12049         PR tree-optimization/82563
12050         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12051         New function.
12052         (graphite_regenerate_ast_isl): Call it.
12053         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12055 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12057         PR tree-optimization/82549
12058         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12059         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12060         bitpos return 0.
12062 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12064         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12065         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12066         _mm_maskz_reduce_ss): New.
12067         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12068         __builtin_ia32_reducess_mask): Ditto..
12069         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12070         * config/i386/sse.md (reduces<mode>): Renamed to ...
12071         (reduces<mode><mask_scalar_name>): ... this.
12072         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12073         Changed to ...
12074         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12075         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12077 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12079         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12080         * selftest-run-tests.c (selftest::run_tests): Call
12081         selftest::unique_ptr_tests_cc_tests.
12082         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12083         * unique-ptr-tests.cc: New file.
12085 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12087         PR sanitizer/82353
12088         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12089         locations.
12090         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12091         (make_hard_regno_born, make_hard_regno_dead): Update
12092         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12094 2017-10-16  Jeff Law  <law@redhat.com>
12096         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12098 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12100         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12102 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12104         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12105         with_cpu if we were configured for an e500v2 target cpu name.
12107 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12109         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12110         * doc/invoke.texi: Document +nodsp as a valid extension for
12111         -mcpu=cortex-m33.
12113 2017-10-16  Martin Liska  <mliska@suse.cz>
12115         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12116         (test_set_range): Likewise.
12117         (test_range_functions): Rename to ...
12118         (test_bit_in_range): ... this.
12119         (sbitmap_c_tests): Add new test.
12121 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12123         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12124         New.
12125         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12126         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12128 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12130         * config/aarch64/aarch64-builtins.c
12131         (aarch64_types_quadopu_lane_qualifiers): New.
12132         (TYPES_QUADOPU_LANE): New.
12133         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12134         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12135         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12136         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12137         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12138         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12139         (Vdottype, DOTPROD): New.
12140         (sur): Add SDOT and UDOT.
12142 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12144         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12145         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12146         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12147         Add TARGET_DOTPROD.
12148         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12149         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12150         Enable TARGET_DOTPROD.
12151         (cortex-a75.cortex-a55): Likewise.
12152         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12154 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12156         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12157         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12158         New.
12159         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12160         New.
12161         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12162         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12163         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12164         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12165         * config/arm/types.md (neon_dot, neon_dot_q): New.
12166         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12168 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12170         * config/arm/arm.h (TARGET_DOTPROD): New.
12171         * config/arm/arm.c (arm_arch_dotprod): New.
12172         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12173         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12174         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12175         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12176         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12177         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12178         * doc/invoke.texi (armv8.2-a): Document dotprod
12180 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12182         * i386.c (ix86_vec_cost): New function.
12183         (ix86_rtx_costs): Handle vector operations better.
12184         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12185         * x86-tune-costs.h: Add new costs to all tables.
12187 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12189         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12190         operations.
12191         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12192         divsd, sqrtss and sqrtsd
12193         * x86-tune-costs.h: Add new entries to all costs.
12194         (znver1_cost): Fix to match real instruction latencies.
12196 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12197             Michael Collison <michael.collison@arm.com>
12199         * compare-elim.c: Include emit-rtl.h.
12200         (can_merge_compare_into_arith): New function.
12201         (try_validate_parallel): Likewise.
12202         (try_merge_compare): Likewise.
12203         (try_eliminate_compare): Call the above when no previous clobber
12204         is available.
12205         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12206         dataflow problems.
12208 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12210         PR middle-end/62263
12211         PR middle-end/82498
12212         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12213         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12215         PR middle-end/62263
12216         PR middle-end/82498
12217         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12218         to be any operand_equal_p operands.  For & (B - 1) require
12219         B to be power of 2.  Recognize
12220         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12222 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12224         PR bootstrap/82553
12225         * optabs.c (expand_memory_blockage): Fix call of
12226         targetm.have_memory_blockage.
12228 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12230         PR bootstrap/82548
12231         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12232         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12233         objects to extra_objs instead of overwriting it.
12235 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12237         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12238         Use any_fp_register_operand as operand[3] predicate.  Simplify
12239         equality test for operands[2] and operands[4] memory location.
12240         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12241         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12242         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12243         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12244         any_fp_register_operand as operand[1] predicate.  Simplify
12245         equality test for operands[0] and operands[3] memory location.
12246         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12247         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12248         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12250 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12252         * target-insns.def: Add memory_blockage.
12253         * optabs.c (expand_memory_blockage): New function.
12254         (expand_asm_memory_barrier): Rename ...
12255         (expand_asm_memory_blockage): ... to this.
12256         (expand_mem_thread_fence): Call expand_memory_blockage
12257         instead of expand_asm_memory_barrier.
12258         (expand_mem_singnal_fence): Ditto.
12259         (expand_atomic_load): Ditto.
12260         (expand_atomic_store): Ditto.
12261         * doc/md.texi (Standard Pattern Names For Generation):
12262         Document memory_blockage instruction pattern.
12264 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12266         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12267         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12268         * config/rl78/rl78.md: New define_expand "adddi3".
12270 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12272         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12273         are set correctly.
12275 2017-10-13  Jeff Law  <law@redhat.com>
12277         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12279 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12281         PR target/82274
12282         * internal-fn.c (expand_mul_overflow): If both operands have
12283         the same highpart of -1 or 0 and the topmost bit of lowpart
12284         is different, overflow is if res <= 0 rather than res < 0.
12286 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12288         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12289         TARGET_P9_VECTOR code for unaligned_load case.
12291 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12293         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12295 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12297         * tree-core.h (tree_contains_struct): Make bool.
12298         * tree.c (tree_contains_struct): Likewise.
12299         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12300         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12301         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12302         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12303         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12305 2017-10-13  Richard Biener  <rguenther@suse.de>
12307         * graphite-isl-ast-to-gimple.c
12308         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12309         parameters and dominance check.
12310         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12311         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12312         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12313         Do not update SSA form here or do intermediate IL verification.
12314         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12315         (graphite_initialize): Remove check on the number of loops in
12316         the function and inline into graphite_transform_loops.
12317         (graphite_finalize): Inline into graphite_transform_loops.
12318         (graphite_transform_loops): Perform SSA update and IL verification
12319         here.
12320         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12322 2017-10-13  Richard Biener  <rguenther@suse.de>
12324         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12325         graphite_expression_type_precision): Avoid global constructor
12326         by moving ...
12327         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12328         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12329         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12330         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12331         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12332         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12334 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12336         PR target/82499
12337         * config/i386/i386.h (ix86_red_zone_size): New.
12338         * config/i386/i386.md (push peephole2s): Replace
12339         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12341 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12342             Alan Hayward  <alan.hayward@arm.com>
12343             David Sherwood  <david.sherwood@arm.com>
12345         * combine.c (can_change_dest_mode): Reject changes in
12346         REGMODE_NATURAL_SIZE.
12348 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12349             Alan Hayward  <alan.hayward@arm.com>
12350             David Sherwood  <david.sherwood@arm.com>
12352         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12353         (expand_debug_source_expr): Likewise.
12354         * combine.c (combine_simplify_rtx): Likewise.
12355         * cse.c (fold_rtx): Likewise.
12356         * fwprop.c (canonicalize_address): Likewise.
12357         * targhooks.c (default_shift_truncation_mask): Likewise.
12359 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12360             Alan Hayward  <alan.hayward@arm.com>
12361             David Sherwood  <david.sherwood@arm.com>
12363         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12364         (widened_mode): Likewise.
12365         (expand_unop): Likewise.
12366         * ree.c (transform_ifelse): Likewise.
12367         (merge_def_and_ext): Likewise.
12368         (combine_reaching_defs): Likewise.
12369         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12371 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12372             Alan Hayward  <alan.hayward@arm.com>
12373             David Sherwood  <david.sherwood@arm.com>
12375         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12376         * combine.c (gen_lowpart_for_combine): Likewise.
12377         * dwarf2out.c (rtl_for_decl_location): Likewise.
12378         * final.c (alter_subreg): Likewise.
12379         * rtlhooks.c (gen_lowpart_general): Likewise.
12380         (gen_lowpart_if_possible): Likewise.
12382 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12383             Alan Hayward  <alan.hayward@arm.com>
12384             David Sherwood  <david.sherwood@arm.com>
12386         * calls.c (expand_call): Use subreg_lowpart_offset.
12387         * cse.c (cse_insn): Likewise.
12388         * regcprop.c (copy_value): Likewise.
12389         (copyprop_hardreg_forward_1): Likewise.
12391 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12393         PR target/82524
12394         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12395         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12396         =Q constraints to +Q and into insn condition add check
12397         that operands[0] and operands[1] are equal.
12398         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12399         =Q constraints to +Q and into insn condition add check
12400         that operands[0] is equal to either operands[1] or operands[2].
12402         PR target/82498
12403         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12404         instead of handling MINUS_EXPR twice (once for each argument),
12405         canonicalize operand order and handle just once, use rtype where
12406         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12408         PR target/82498
12409         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12410         any values of __C while still being pattern recognizable as a simple
12411         rotate instruction.
12413 2017-10-13  Richard Biener  <rguenther@suse.de>
12415         PR tree-optimization/82451
12416         Revert
12417         2017-10-02  Richard Biener  <rguenther@suse.de>
12419         PR tree-optimization/82355
12420         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12421         a mapping for the enclosing loop but avoid generating one for
12422         the loop tree root.
12423         (copy_bb_and_scalar_dependences): Remove premature codegen
12424         error on PHIs in blocks duplicated into multiple places.
12425         * graphite-scop-detection.c
12426         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12427         in the region use it as loop and nest to analyze the DR in.
12428         (try_generate_gimple_bb): Likewise.
12429         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12430         (add_loop_constraints): For blocks in a loop not in the region
12431         create a dimension with a single iteration.
12432         * sese.h (gbb_loop_at_index): Remove assert.
12434         * cfgloop.c (loop_preheader_edge): For the loop tree root
12435         return the single successor of the entry block.
12436         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12437         Reset the SCEV hashtable and niters.
12438         * graphite-scop-detection.c
12439         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12440         assert that we only have POLYNOMIAL_CHREC that vary in loops
12441         contained in the region.
12442         (scop_detection::graphite_can_represent_expr): Adjust.
12443         (scop_detection::stmt_has_simple_data_refs_p): For loops
12444         not in the region set loop to NULL.  The nest is now the
12445         entry edge to the region.
12446         (try_generate_gimple_bb): Likewise.
12447         * sese.c (scalar_evolution_in_region): Adjust for
12448         instantiate_scev change.
12449         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12450         Make nest parameter the edge into the region.
12451         (create_data_ref): Likewise.
12452         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12453         entry edge into a region and adjust instantiate_scev calls.
12454         (create_data_ref): Likewise.
12455         (graphite_find_data_references_in_stmt): Likewise.
12456         (find_data_references_in_stmt): Pass the loop preheader edge
12457         from the nest argument.
12458         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12459         parameter the edge into the region.
12460         (instantiate_parameters): Use the loop preheader edge as entry.
12461         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12462         NULL loop.
12463         (get_instantiated_value_entry): Make instantiate_below parameter
12464         the edge into the region.
12465         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12466         when we cannot use loop-based instantiation instantiate by
12467         walking use-def chains.
12468         (instantiate_scev_poly): Adjust.
12469         (instantiate_scev_binary): Likewise.
12470         (instantiate_scev_convert): Likewise.
12471         (instantiate_scev_not): Likewise.
12472         (instantiate_array_ref): Remove.
12473         (instantiate_scev_3): Likewise.
12474         (instantiate_scev_2): Likewise.
12475         (instantiate_scev_1): Likewise.
12476         (instantiate_scev_r): Do not blindly handle N-operand trees.
12477         Do not instantiate array-refs.  Handle all constants and invariants.
12478         (instantiate_scev): Make instantiate_below parameter
12479         the edge into the region.
12480         (resolve_mixers): Use the loop preheader edge for the region
12481         parameter to instantiate_scev_r.
12482         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12484 2017-10-13  Richard Biener  <rguenther@suse.de>
12486         PR tree-optimization/82525
12487         * graphite-isl-ast-to-gimple.c
12488         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12489         out from ...
12490         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12491         Fail code generation when we cannot represent the isl integer.
12492         (binary_op_to_tree): Elide modulo operations that are no-ops
12493         in the type we code generate.  Remove now superfluous code
12494         generation errors.
12496 2017-10-13  Richard Biener  <rguenther@suse.de>
12498         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12499         (scop_detection::harmful_loop_in_region): Remove premature
12500         IV type restriction.
12501         (scop_detection::graphite_can_represent_scev): We can handle
12502         pointer IVs just fine.
12504 2017-10-13  Alan Modra  <amodra@gmail.com>
12506         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12507         "Clobbers and Scratch Registers".  Add paragraph on
12508         alternative to clobbers for scratch registers and OpenBLAS
12509         example.
12511 2017-10-13  Alan Modra  <amodra@gmail.com>
12513         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12514         example of a memory input for a string of known length.  Move
12515         commentary out of table.  Add a number of new examples
12516         covering array memory inputs.
12518 2017-10-12  Martin Liska  <mliska@suse.cz>
12520         PR tree-optimization/82493
12521         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12522         (test_range_functions): New function.
12523         (sbitmap_c_tests): Likewise.
12524         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12525         * selftest.h (sbitmap_c_tests): New function.
12527         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12529 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12531         * config/rs6000/amo.h: Fix spacing issue.
12533 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12535         PR target/82498
12536         * config/i386/i386.md (*ashl<mode>3_mask_1,
12537         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12538         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12539         patterns.
12541 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12543         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12544         (profile_probability): Set max_probability
12545         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12546         in temporaries.
12547         * profile-count.c (profile_probability::differs_from_p): Do not
12548         rely on max_probaiblity == 10000
12550 2017-10-12  Jeff Law  <law@redhat.com>
12552         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12553         negative offsets.
12555 2017-10-12  Martin Sebor  <msebor@redhat.com>
12557         PR other/82301
12558         PR c/82435
12559         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12560         (handle_alias_pairs): Call it.
12561         * common.opt (-Wattribute-alias): New option.
12562         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12563         * doc/invoke.texi (-Wattribute-alias): Document.
12565 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12567         Revert
12568         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12569         PR sanitizer/82353
12570         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12571         locations.
12572         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12573         (make_hard_regno_born, make_hard_regno_dead): Update
12574         bb_killed_pseudos and bb_gen_pseudos.
12576 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12578         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12580 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12582         * config/alpha/alpha.c (alpha_split_conditional_move):
12583         Use std::swap instead of manually swapping.
12584         (alpha_stdarg_optimize_hook): Ditto.
12585         (alpha_canonicalize_comparison): Ditto.
12587 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12589         * tree-loop-distribution.c (struct builtin_info): New struct.
12590         (struct partition): Refactor fields into struct builtin_info.
12591         (partition_free): Free struct builtin_info.
12592         (build_size_arg_loc, build_addr_arg_loc): Delete.
12593         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12594         information from struct builtin_info.
12595         (find_single_drs): New function refactored from classify_partition.
12596         Also moved builtin validity checks to this function.
12597         (compute_access_range, alloc_builtin): New functions.
12598         (classify_builtin_st, classify_builtin_ldst): New functions.
12599         (classify_partition): Refactor code into functions find_single_drs,
12600         classify_builtin_st and classify_builtin_ldst.
12601         (distribute_loop): Don't do runtime alias check when distributing
12602         loop nest.
12603         (find_seed_stmts_for_distribution): New function.
12604         (pass_loop_distribution::execute): Refactor code finding seed
12605         stmts into above function.  Support distribution for the innermost
12606         two-level loop nest.  Adjust dump information.
12608 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12610         * tree-loop-distribution.c: Adjust the general comment.
12611         (NUM_PARTITION_THRESHOLD): New macro.
12612         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12613         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12614         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12615         in call to build_partition_graph.
12616         (finalize_partitions): New parameter.  Make loop distribution more
12617         conservative by fusing more partitions.
12618         (distribute_loop): Don't do runtime alias check in case of loop nest
12619         distribution.
12620         (find_seed_stmts_for_distribution): New function.
12621         (prepare_perfect_loop_nest): New function.
12622         (pass_loop_distribution::execute): Refactor code finding seed stmts
12623         and loop nest into above functions.  Support loop nest distribution.
12624         Adjust dump information accordingly.
12626 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12628         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12629         and set PTYPE_SEQUENTIAL for merged partition.
12631 2017-10-12  Richard Biener  <rguenther@suse.de>
12633         PR tree-optimization/69728
12634         Revert
12635         2017-09-19  Richard Biener  <rguenther@suse.de>
12637         PR tree-optimization/69728
12638         * graphite-sese-to-poly.c (schedule_error): New global.
12639         (add_loop_schedule): Handle empty domain by failing the
12640         schedule.
12641         (build_original_schedule): Handle schedule_error.
12643         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12644         domain by returning an unchanged schedule.
12646 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12648         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12649         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12651 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12653         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12654         Handle params.def.
12656 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12658         PR c++/82159
12659         * expr.c (store_field): Don't optimize away bitsize == 0 store
12660         from CALL_EXPR with addressable return type.
12662 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12664         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12665         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12666         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12667         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12668         TARGET_ISEL instead of TARGET_ISEL<sel>.
12670 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12672         * config/rs6000/rs6000.c
12673         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12675 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12677         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12678         Move up in file.
12679         (reg_or_cint_operand): Fix comment.
12680         (reg_or_zero_operand): New predicate.
12681         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12682         * config/rs6000/rs6000.c (output_isel): Delete.
12683         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12684         instead of reg_or_cint_operand.  Output instruction directly (not via
12685         output_isel).
12686         (isel_unsigned_<mode>): Ditto.
12687         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12688         gpc_reg_operand.  Add an instruction alternative for this.  Output
12689         instruction directly.
12690         (*isel_reversed_unsigned_<mode>): Ditto.
12692 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12694         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12695         (TARGET_CANONICALIZE_COMPARISON): Define.
12697 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12699         PR target/81422
12700         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12701         Check whether the dest is REG before adding REG_EQUIV note.
12703 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12705         PR sanitizer/82353
12706         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12707         locations.
12708         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12709         (make_hard_regno_born, make_hard_regno_dead): Update
12710         bb_killed_pseudos and bb_gen_pseudos.
12712 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12714         * incpath.h (enum incpath_kind): Name enum, prefix values.
12715         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12716         * incpath.c (heads, tails): Use INC_MAX.
12717         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12718         (merge_include_chains, split_quote_chain,
12719         register_include_chains): Update incpath_kind names.
12720         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12721         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12722         names.
12723         (add_framework_path, darwin_register_objc_includes): Likewise.
12724         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12726 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12728         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12729         Do not use float_operator operator predicate.
12730         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12731         * config/i386/predicates.md (float_operator): Remove predicate.
12733 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12735         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12736         (*jccxf_i387): Ditto.
12737         (*jcc<mode>_i387): Ditto.
12738         (*jccu<mode>_i387): Ditto.
12739         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12740         (*jcc_*_i387 splitters): Remove.
12741         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12742         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12743         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12744         Remove predicate.
12746 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12748         * profile-count.h (slow_safe_scale_64bit): New function.
12749         (safe_scale_64bit): New inline.
12750         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12751         * profile-count.c: Include wide-int.h
12752         (slow_safe_scale_64bit): New.
12754 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12756         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12757         HAS_DECL_ASSEMBLER_NAME_P.
12758         * gimple-expr.c (gimple_decl_printable_name: Check
12759         HAS_DECL_ASSEMBLER_NAME_P too.
12760         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12761         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12762         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12763         * passes.c (rest_of_decl_compilation): Check
12764         HAS_DECL_ASSEMBLER_NAME_P too.
12765         * recog.c (verify_changes): Likewise.
12766         * tree-pretty-print.c (dump_decl_name): Likewise.
12767         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12769         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12770         (SET_DECL_ASSEMBLER_NAME): Use it.
12771         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12772         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12773         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12775 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12777         * config.gcc (i386, x86_64): Add extra objects.
12778         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12779         (ix86_min_insn_size): Declare.
12780         (ix86_issue_rate): Declare.
12781         (ix86_adjust_cost): Declare.
12782         (ia32_multipass_dfa_lookahead): Declare.
12783         (ix86_macro_fusion_p): Declare.
12784         (ix86_macro_fusion_pair_p): Declare.
12785         (ix86_bd_has_dispatch): Declare.
12786         (ix86_bd_do_dispatch): Declare.
12787         (ix86_core2i7_init_hooks): Declare.
12788         (ix86_atom_sched_reorder): Declare.
12789         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12790         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12791         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12792         (rip_relative_addr_p): Rename to ...
12793         (ix86_rip_relative_addr_p): ... this one; export.
12794         (memory_address_length): Update.
12795         (ix86_issue_rate): Move to x86-tune-sched.c.
12796         (ix86_flags_dependent): Move to x86-tune-sched.c.
12797         (ix86_agi_dependent): Move to x86-tune-sched.c.
12798         (exact_dependency_1): Move to x86-tune-sched.c.
12799         (exact_store_load_dependency): Move to x86-tune-sched.c.
12800         (ix86_adjust_cost): Move to x86-tune-sched.c.
12801         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12802         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12803         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12804         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12805         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12806         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12807         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12808         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12809         (min_insn_size): Rename to ...
12810         (ix86_min_insn_size): ... this one; export.
12811         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12812         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12813         (core2i7_first_cycle_multipass_backtrack): Move to
12814         x86-tune-sched-core.c.
12815         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12816         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12817         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12818         (ix86_avoid_jump_mispredicts): Update.
12819         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12820         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12821         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12822         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12823         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12824         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12825         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12826         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12827         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12828         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12829         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12830         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12831         (BIG): Move to ix86-tune-sched-bd.c.
12832         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12833         (enum insn_path): Move to ix86-tune-sched-bd.c.
12834         (get_mem_group): Move to ix86-tune-sched-bd.c.
12835         (is_cmp): Move to ix86-tune-sched-bd.c.
12836         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12837         (is_branch): Move to ix86-tune-sched-bd.c.
12838         (is_prefetch): Move to ix86-tune-sched-bd.c.
12839         (init_window): Move to ix86-tune-sched-bd.c.
12840         (allocate_window): Move to ix86-tune-sched-bd.c.
12841         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12842         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12843         (process_end_window): Move to ix86-tune-sched-bd.c.
12844         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12845         (find_constant): Move to ix86-tune-sched-bd.c.
12846         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12847         (has_immediate): Move to ix86-tune-sched-bd.c.
12848         (get_insn_path): Move to ix86-tune-sched-bd.c.
12849         (get_insn_group): Move to ix86-tune-sched-bd.c.
12850         (count_num_restricted): Move to ix86-tune-sched-bd.c.
12851         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12852         (add_insn_window): Move to ix86-tune-sched-bd.c.
12853         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12854         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12855         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12856         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12857         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12858         (do_dispatch): Move to ix86-tune-sched-bd.c.
12859         (has_dispatch): Move to ix86-tune-sched-bd.c.
12860         * config/i386/t-i386: Add new object files.
12861         * config/i386/x86-tune-costs.h: New file.
12862         * config/i386/x86-tune-sched-atom.c: New file.
12863         * config/i386/x86-tune-sched-bd.c: New file.
12864         * config/i386/x86-tune-sched-core.c: New file.
12865         * config/i386/x86-tune-sched.c: New file.
12867 2017-10-11  Liu Hao  <lh_mouse@126.com>
12869         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
12870         the generic version below instead.
12871         (should_colorize): Recognize Windows consoles as terminals
12872         for MinGW targets.
12873         * pretty-print.c [__MINGW32__] (write_all): New function.
12874         [__MINGW32__] (find_esc_head): Likewise.
12875         [__MINGW32__] (find_esc_terminator): Likewise.
12876         [__MINGW32__] (eat_esc_sequence): Likewise.
12877         [__MINGW32__] (mingw_ansi_fputs): New function that handles
12878         ANSI escape codes.
12879         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
12880         for MinGW targets.
12882 2017-10-11  Richard Biener  <rguenther@suse.de>
12884         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
12885         Properly call analyze_scalar_evolution with the loop of the stmt.
12887 2017-10-11  Richard Biener  <rguenther@suse.de>
12889         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
12890         * tree-core.h (tree_base): Add chrec_var union member.
12891         * tree.h (CHREC_VAR): Remove.
12892         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
12893         * tree-chrec.h (build_polynomial_chrec): Adjust.
12894         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
12895         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
12897 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
12899         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
12900         * match.pd: ... here.
12901         ((T) X == (T) Y): Relax condition.
12903 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
12905         PR tree-optimization/82472
12906         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
12907         comment.
12908         (break_alias_scc_partitions): Update postorder number.
12910 2017-10-11  Martin Liska  <mliska@suse.cz>
12912         PR sanitizer/82490
12913         * opts.c (parse_no_sanitize_attribute): Do not use error_value
12914         variable.
12915         * opts.h (parse_no_sanitize_attribute): Remove last argument.
12917 2017-10-11  Martin Liska  <mliska@suse.cz>
12919         * print-rtl.c (print_insn): Move declaration of idbuf
12920         to same scope as name.
12922 2017-10-11  Martin Liska  <mliska@suse.cz>
12924         Revert r253637:
12926         PR sanitizer/82484
12927         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12928         volatile arguments.
12930 2017-10-11  Martin Liska  <mliska@suse.cz>
12932         PR sanitizer/82484
12933         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12934         volatile arguments.
12936 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
12938         * config.gcc (default_gnu_indirect_function): Default to yes for
12939         arm*-*-linux* with glibc.
12941 2017-10-11  Richard Biener  <rguenther@suse.de>
12943         * tree-scalar-evolution.c (get_scalar_evolution): Handle
12944         default-defs and types we do not want to analyze.
12945         (interpret_loop_phi): Replace unreachable code with an assert.
12946         (compute_scalar_evolution_in_loop): Remove and inline ...
12947         (analyze_scalar_evolution_1): ... here, replacing condition with
12948         what makes the intent clearer.  Remove handling of cases
12949         get_scalar_evolution now handles.
12951 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
12953         PR rtl-optimization/81434
12954         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
12955         comment for main loop.  In sched_group_found if, also add checks for
12956         pass and min_cost_group.
12958 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
12960         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
12961         (rs6000_insn_cost): New function.
12962         * config/rs6000/rs6000.md (cost): New attribute.
12964 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
12965             H.J. Lu  <hongjiu.lu@intel.com>
12967         PR target/79565
12968         PR target/82483
12969         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
12970         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
12971         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
12972         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
12973         to OPTION_MASK_ISA_AVX512VL - builtins that have both
12974         OPTION_MASK_ISA_MMX and some other bit set require both
12975         mmx and the ISAs without the mmx bit.
12976         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
12977         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
12978         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
12979         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
12980         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
12981         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
12982         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
12983         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
12984         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
12985         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
12986         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
12987         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
12988         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
12989         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
12990         Add OPTION_MASK_ISA_MMX.
12992 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
12994         * config.gcc (armv7*-*-freebsd*): New target.
12995         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
12997 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
12999         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13000         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13001         spot in the file.
13003 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13005         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13006         a template parameter.
13007         (WIDE_INT_REF_FOR): Update accordingly.
13008         * tree.h (wi::int_traits <const_tree>): Delete.
13009         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13010         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13011         (wi::tree_to_wide_ref): New typedef.
13012         (wi::to_wide): New function.
13013         * calls.c (get_size_range): Use wi::to_wide when operating on
13014         trees as wide_ints.
13015         * cgraph.c (cgraph_node::create_thunk): Likewise.
13016         * config/i386/i386.c (ix86_data_alignment): Likewise.
13017         (ix86_local_alignment): Likewise.
13018         * dbxout.c (stabstr_O): Likewise.
13019         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13020         * expr.c (const_vector_from_tree): Likewise.
13021         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13022         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13023         (fold_negate_expr_1, int_const_binop_1, const_binop)
13024         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13025         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13026         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13027         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13028         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13029         (fold_not_const, round_up_loc): Likewise.
13030         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13031         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13032         (alloca_call_type): Likewise.
13033         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13034         * godump.c (go_output_typedef): Likewise.
13035         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13036         * internal-fn.c (get_min_precision): Likewise.
13037         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13038         * ipa-polymorphic-call.c
13039         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13040         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13041         (ipa_modify_call_arguments): Likewise.
13042         * match.pd: Likewise.
13043         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13044         * print-tree.c (print_node_brief, print_node): Likewise.
13045         * stmt.c (expand_case): Likewise.
13046         * stor-layout.c (layout_type): Likewise.
13047         * tree-affine.c (tree_to_aff_combination): Likewise.
13048         * tree-cfg.c (group_case_labels_stmt): Likewise.
13049         * tree-data-ref.c (dr_analyze_indices): Likewise.
13050         (prune_runtime_alias_test_list): Likewise.
13051         * tree-dump.c (dequeue_and_dump): Likewise.
13052         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13053         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13054         * tree-pretty-print.c (dump_generic_node): Likewise.
13055         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13056         (simple_iv_with_niters): Likewise.
13057         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13058         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13059         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13060         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13061         (refine_value_range_using_guard, number_of_iterations_ne_max)
13062         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13063         (get_cst_init_from_scev, record_nonwrapping_iv)
13064         (scev_var_range_cant_overflow): Likewise.
13065         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13066         * tree-ssa-pre.c (compute_avail): Likewise.
13067         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13068         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13069         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13070         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13071         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13072         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13073         (dump_case_nodes, try_switch_expansion): Likewise.
13074         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13075         (vect_do_peeling): Likewise.
13076         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13077         * tree-vect-stmts.c (vectorizable_load): Likewise.
13078         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13079         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13080         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13081         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13082         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13083         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13084         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13085         (evrp_dom_walker::before_dom_children): Likewise.
13086         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13087         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13088         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13089         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13090         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13091         (get_range_pos_neg): Likewise.
13092         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13093         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13094         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13095         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13096         * config/avr/avr.c (avr_fold_builtin): Likewise.
13097         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13098         * config/msp430/msp430.c (msp430_attr): Likewise.
13099         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13100         * config/powerpcspe/powerpcspe-c.c
13101         (altivec_resolve_overloaded_builtin): Likewise.
13102         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13103         (rs6000_expand_ternop_builtin): Likewise.
13104         * config/rs6000/rs6000-c.c
13105         (altivec_resolve_overloaded_builtin): Likewise.
13106         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13107         (rs6000_expand_ternop_builtin): Likewise.
13108         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13110 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13112         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13113         when copying loop nest with only one inner loop.
13115 2017-10-10  Richard Biener  <rguenther@suse.de>
13117         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13118         blocks if SCEV is active.
13119         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13120         dead code.
13121         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13122         (scev_initialize): Assert we are not yet initialized.
13124 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13126         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13127         inner loop's exit stmt by making it always exit the loop, otherwise
13128         we would generate an infinite empty loop.
13130 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13132         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13133         renaming variables in new preheader if it's deleted.
13135 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13137         * tree-loop-distribution.c (struct partition): Remove unused field
13138         loops of the structure.
13139         (partition_alloc, partition_free): Ditto.
13140         (build_rdg_partition_for_vertex): Ditto.
13142 2017-10-09  Jeff Law  <law@redhat.com>
13144         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13145         return type to match prototype and documentation.
13147 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13149         * config/rs6000/rs6000.c (processor_costs): Move to ...
13150         * config/rs6000/rs6000.h: ... here.
13151         (rs6000_cost): Declare.
13153 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13155         * except.c (setjmp_fn): New global variable.
13156         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13157         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13158         if DONT_USE_BUILTIN_SETJMP is defined.
13160 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13162         * target.def (insn_cost): New hook.
13163         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13164         * doc/tm.texi: Regenerate.
13165         * rtlanal.c (insn_cost): Use the new hook.
13167 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13169         * combine.c (combine_validate_cost): Compute the new insn_cost,
13170         not just pattern_cost.
13171         (try_combine): Adjust comment.
13173 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13175         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13176         insn_cost.
13177         * combine.c (uid_insn_cost): Adjust comment.
13178         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13179         of insn_rtx_cost
13180         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13181         * dse.c (find_shift_sequence): Ditto.
13182         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13183         (bb_valid_for_noce_process_p): Use pattern_cost.
13184         * rtl.h (insn_rtx_cost): Delete.
13185         (pattern_cost): New prototype.
13186         (insn_cost): New prototype.
13187         * rtlanal.c (insn_rtx_cost): Rename to...
13188         (pattern_cost): ... this.
13189         (insn_cost): New.
13191 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13193         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13194         (*jcc<mode>_0_r_i387): Ditto.
13195         (*jccxf_r_i387): Ditto.
13196         (*jcc<mode>_r_i387): Ditto.
13197         (*jccu<mode>_r_i387): Ditto.
13198         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13199         (*jcc): Rename from *jcc_1.
13201 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13203         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13204         deferred rescans after the lvx/stvx recombination pre-pass.
13206 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13208         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13209         memory operation instruction support.
13210         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13211         (rs6000-ibm-aix[789]*): Likewise.
13212         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13213         Document new functions.
13215 2017-10-09  Richard Biener  <rguenther@suse.de>
13217         PR tree-optimization/82397
13218         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13219         equality only for semantically equal trees.
13221 2017-10-09  Richard Biener  <rguenther@suse.de>
13223         PR tree-optimization/82449
13224         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13225         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13226         allow constant addresses.
13227         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13228         are linear.
13230 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13232         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13233         flags.
13235 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13237         PR target/82463
13238         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13239         definitions.
13241 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13243         PR target/82465
13244         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13246 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13248         PR target/82464
13249         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13250         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13252 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13254         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13255         (WI_BINARY_PREDICATE_RESULT): Likewise.
13256         (wi::binary_traits::operator_result): New type.
13257         (wi::binary_traits::predicate_result): Likewise.
13258         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13259         (generic_wide_int::operator==, generic_wide_int::operator!=)
13260         (generic_wide_int::operator&, generic_wide_int::and_not)
13261         (generic_wide_int::operator|, generic_wide_int::or_not)
13262         (generic_wide_int::operator^, generic_wide_int::operator+
13263         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13264         Delete.
13265         (operator~, unary operator-, operator==, operator!=, operator&)
13266         (operator|, operator^, operator+, binary operator-, operator*): New
13267         functions.
13268         * expr.c (get_inner_reference): Use wi::bit_and_not.
13269         * fold-const.c (fold_binary_loc): Likewise.
13270         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13271         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13272         (bit_value_binop): Likewise.
13273         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13274         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13275         (extract_range_from_binary_expr_1): Likewise.
13276         (masked_increment): Likewise.
13277         (simplify_bit_ops_using_ranges): Likewise.
13279 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13281         PR hsa/82416
13282         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13283         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13284         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13285         COMPLEX types.
13286         (hsa_fixup_mov_insn_type): New function.
13287         (hsa_op_with_type::get_in_type): Use it.
13288         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13289         immediates in an assert.
13290         (hsa_op_with_type::extend_int_to_32bit): New method.
13291         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13292         types.  Convert to dest type if necessary.
13293         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13294         (reg_for_gimple_ssa): Pass false as min32int to
13295         hsa_type_for_scalar_tree_type.
13296         (gen_hsa_addr): Fixup type when creating addresable temporary.
13297         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13298         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13299         necessary.  Call hsa_fixup_mov_insn_type.
13300         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13301         extend operands and convert to dest type if necessary.
13302         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13303         to dest type if necessary.
13304         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13305         if conversion nt necessary and size matches.
13306         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13307         to dest type if necessary.
13308         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13309         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13310         necessary.
13311         (gen_hsa_clrsb): Likewise.
13312         (gen_hsa_ffs): Likewise.
13313         (gen_hsa_divmod): Extend operands and convert to dest type if
13314         necessary.
13315         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13317 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13319         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13320         Remove empty default arguments.  Use a brace block as output
13321         statement.
13322         (conditional return): Ditto.
13323         (jump): Ditto.
13324         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13325         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13326         Ditto.
13327         (group_ending_nop): Ditto.
13328         (doloop_end): Ditto.
13329         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13330         (splitters for those): Ditto.
13332 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13334         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13335         a conditional jump (and the compare for it) so that pc_rtx is the
13336         last operand.
13337         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13338         for the deleted and renamed ctr<mode>_internal[234] patterns.
13339         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13340         Delete second conditional return pattern.
13341         (ctr<mode>_internal2): Delete this second bdnz pattern.
13342         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13343         (ctr<mode>_internal4): Delete this second bdz pattern.
13345 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13347         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13348         (always_initialized_rtx_for_ssa_name_p): New predicate.
13349         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13350         (finish_out_of_ssa): Free new field of SA.
13351         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13352         * tree-ssa-coalesce.c: Include tree-ssa.h.
13353         (get_parm_default_def_partitions): Remove extern keyword.
13354         (get_undefined_value_partitions): New function.
13355         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13356         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13357         uninitialized bits.
13358         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13360 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13362         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13364 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13366         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13367         for targets that preffer 128bit.
13369 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13371         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13373 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13375         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13376         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13377         prevent DSE.
13378         (thumb_set_return_address): Likewise.
13380 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13382         * common/config/arm/arm-common.c (arm_except_unwind_info):
13383         Handle DWARF2_UNWIND_INFO.
13385 2017-10-07  Michael Collison <michael.collison@arm.com>
13387         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13388         New pattern.
13390 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13392         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13393         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13394         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13395         defined, force the creation of a new block for a dispatch label.
13397 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13399         * invoke.texi (Wsuggest-attribute=cold): Document.
13400         * common.opt (Wsuggest-attribute=cold): New
13401         * ipa-pure-const.c (warn_function_cold): New function.
13402         * predict.c (compute_function_frequency): Use it.
13403         * predict.h (warn_function_cold): Declare.
13405 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13407         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13408         Update profile.
13410 2017-10-06  Martin Liska  <mliska@suse.cz>
13412         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13413         keyword for member functions.
13414         (struct sanopt_tree_couple): New struct.
13415         (struct sanopt_tree_couple_hash): New function.
13416         (struct sanopt_ctx): Add new hash_map.
13417         (has_dominating_ubsan_ptr_check): New function.
13418         (record_ubsan_ptr_check_stmt): Likewise.
13419         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13420         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13421         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13423 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13425         PR target/82440
13426         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13427         aarch64_simd_valid_immediate on CONST_VECTORs.
13428         (aarch64_reg_or_bic_imm): Likewise.
13430 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13432         PR rtl-optimization/82396
13433         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13435 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13437         * graphite-dependences.c (scop_get_reads): Move code to...
13438         (scop_get_must_writes): Move code to...
13439         (scop_get_may_writes): Move code to...
13440         (scop_get_reads_and_writes): ... here.
13441         (scop_get_dependences): Call scop_get_reads_and_writes.
13443 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13445         PR tree-optimization/82434
13446         * fold-const.h (can_native_encode_type_p,
13447         can_native_encode_string_p): Remove.
13448         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13449         don't encode anything, just return what would be otherwise returned.
13450         (native_encode_fixed, native_encode_complex, native_encode_vector):
13451         Likewise.
13452         (native_encode_string): Likewise.  Inline by hand
13453         can_native_encode_string_p.
13454         (can_native_encode_type_p): Remove.
13455         (can_native_encode_string_p): Remove.
13456         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13457         STRING_CSTs using can_native_encode_string_p, test all
13458         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13459         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13460         argument from native_encode_expr.
13461         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13462         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13463         but just 2.
13465 2017-10-06  Richard Biener  <rguenther@suse.de>
13467         PR tree-optimization/82397
13468         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13469         operand_equal_p but rely on data_ref_compare_tree for detecting
13470         equalities.
13471         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13472         to match up with dr_group_sort_cmp.
13474 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13476         PR target/82322
13477         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13478         builtin.
13479         * config/s390/s390-builtin-types.def: Regenerate.
13481 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13483         PR target/82317
13484         * config/s390/s390-builtin-types.def: Regenerate.
13485         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13486         Change flag from B_VXE to B_VX.
13487         (s390_vec_min_dbl): Remove B_VXE flag.
13489 2017-10-06  Richard Biener  <rguenther@suse.de>
13491         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13492         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13493         translate_isl_ast_to_gimple::is_valid_rename,
13494         translate_isl_ast_to_gimple::get_rename,
13495         translate_isl_ast_to_gimple::get_def_bb_for_const,
13496         translate_isl_ast_to_gimple::get_new_name,
13497         translate_isl_ast_to_gimple::collect_all_ssa_names,
13498         translate_isl_ast_to_gimple::copy_loop_phi_args,
13499         translate_isl_ast_to_gimple::collect_all_ssa_names,
13500         translate_isl_ast_to_gimple::copy_loop_phi_args,
13501         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13502         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13503         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13504         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13505         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13506         translate_isl_ast_to_gimple::copy_cond_phi_args,
13507         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13508         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13509         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13510         translate_isl_ast_to_gimple::rename_uses,
13511         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13512         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13513         (set_rename_for_each_def): Likewise.
13514         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13515         here.  Handle rewriting SCEV analyzable uses here.
13516         (copy_bb_and_scalar_dependences): Generate code for PHI
13517         copy-in/outs.
13518         (graphite_regenerate_ast_isl): Adjust.
13519         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13520         (add_write, add_read): New functions.
13521         (build_cross_bb_scalars_def): Use it and simplify.
13522         (build_cross_bb_scalars_use): Likewise.
13523         (graphite_find_cross_bb_scalar_vars): Inline into...
13524         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13525         simulating out-of-SSA.  Compute liveout and add dependencies.
13526         (build_scops): Force an empty entry block.
13527         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13528         members.
13529         (sese_build_liveouts): Declare.
13530         (sese_trivially_empty_bb_p): Likewise.
13531         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13532         compute liveout and debug_liveout.
13533         (sese_bad_liveouts_use): Remove.
13534         (sese_reset_debug_liveouts_bb): Likewise.
13535         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13536         (sese_build_liveouts): Build liveout and debug_liveout and store
13537         it in region.
13538         (new_sese_info): Adjust.
13539         (free_sese_info): Likewise.
13540         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13541         do not build liveout here.
13542         (move_sese_in_condition): Adjust region entry.
13543         (scev_analyzable_p): Match up with chrec_apply requirements.
13544         (sese_trivially_empty_bb_p): New.
13545         * tree-into-ssa.c (get_reaching_def): Properly support generating
13546         default-defs for incremental rewrite of anonymous names.
13548 2017-10-06  Richard Biener  <rguenther@suse.de>
13550         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13551         precision do not perform modulo reduction.
13553 2017-10-06  Richard Biener  <rguenther@suse.de>
13555         PR tree-optimization/82436
13556         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13557         conservatively choose the vectorization factor when checking
13558         whether we can perform the required load permutation.
13559         (vect_transform_slp_perm_load): Assert when we may not fail.
13561 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13563         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13564         message for incompatible -msdata=* and -mcall-* options.
13566 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13568         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13569         rate for post-reload scheduling.
13571 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13573         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13575 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13577         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13578         to improve monte carlo in scimark.
13580 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13582         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13583         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13584         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13585         pentium4_cost, nocona_cost): Set reassociation width to 1.
13586         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13587         width to 2 for fp operations and 1 otherwise.
13588         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13589         for int and fp.
13590         (atom_cost): Set reassociation width to 2.
13591         (slm_cost, generic_cost): Set fp reassociation width
13592         to 2 and 1 otherwise.
13593         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13594         (core_cost): Set fp reassociation width to 4 and vector to 2.
13595         (ix86_reassociation_width): Rewrite using cost table; special case
13596         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13597         and TARGET_AVX128_OPTIMAL.
13598         * config/i386/i386.h (processor_costs): Add
13599         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13600         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13601         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13602         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13603         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13604         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13606 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13608         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13610 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13612         * config/arm/arm.c (arm_test_fpu_data): New.
13613         (arm_run_selftests): Call arm_test_fpu_data.
13615 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13617         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13618         decl.
13619         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13621 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13623         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13624         check type for aarch64_simd_valid_immediate.
13625         (aarch64_output_simd_mov_immediate): Update prototype.
13626         (aarch64_simd_valid_immediate): Update prototype.
13627         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13628         support for ORR-immediate.
13629         (and<mode>3): modified pattern to add support for BIC-immediate.
13630         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13631         now checks for valid immediate for BIC and ORR based on new enum
13632         argument.
13633         (aarch64_output_simd_mov_immediate): Function now used to output
13634         BIC/ORR imm as well based on new enum argument.
13635         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13636         (Db) : Likewise.
13637         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13638         (aarch64_reg_or_bic_imm): Likewise.
13640 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13642         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13643         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13645 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13647         Revert r253399:
13649         PR rtl-optimization/82396
13650         * haifa-sched.c (autopref_multipass_init): Simplify
13651         initialization.
13652         (autopref_rank_data): Simplify sort order.
13653         * sched-int.h (autopref_multipass_data_): Remove
13654         multi_mem_insn_p, min_offset and max_offset.
13656 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13658         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13660 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13662         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13663         vect_doubleint_cvt.
13665 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13667         * doc/sourcebuild.texi: Document vect_long_mult.
13669 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13671         PR tree-optimization/82413
13672         * fold-const.c (build_range_check): Use widest_int when comparing
13673         the maximum ETYPE value with HIGH.
13675 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13677         PR rtl-optimization/82396
13678         * haifa-sched.c (autopref_multipass_init): Simplify
13679         initialization.
13680         (autopref_rank_data): Simplify sort order.
13681         * sched-int.h (autopref_multipass_data_): Remove
13682         multi_mem_insn_p, min_offset and max_offset.
13684 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13686         PR tree-optimization/82381
13687         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13688         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13689         is not.
13691         PR tree-optimization/82374
13692         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13693         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13694         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13695         current_function_decl to the new decl.
13697 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13699         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13700         helper macro for IEEE float128 hardware built-in functions.
13701         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13702         semantics.
13703         (TRUNCF128_ODD): Likewise.
13704         (ADDF128_ODD): Likewise.
13705         (SUBF128_ODD): Likewise.
13706         (MULF128_ODD): Likewise.
13707         (DIVF128_ODD): Likewise.
13708         (FMAF128_ODD): Likewise.
13709         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13710         UNSPEC_TRUNC_ROUND_TO_ODD.
13711         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13712         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13713         floating point round to odd instructions.
13714         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13715         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13716         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13717         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13718         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13719         (trunc<mode>sf2_hw): Change the truncate with round to odd
13720         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13721         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13722         to odd hardware instructions.
13723         (sub<mode>3_odd): Likewise.
13724         (mul<mode>3_odd): Likewise.
13725         (div<mode>3_odd): Likewise.
13726         (sqrt<mode>2_odd): Likewise.
13727         (fma<mode>4_odd): Likewise.
13728         (fms<mode>4_odd): Likewise.
13729         (nfma<mode>4_odd): Likewise.
13730         (nfms<mode>4_odd): Likewise.
13731         (trunc<mode>df2_odd): Change the truncate with round to odd
13732         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13733         function.
13734         * doc/extend.texi (PowerPC built-in functions): Update documentation
13735         for existing IEEE float128-bit built-in functions.  Add built-in
13736         functions that generate the IEEE 128-bit floating point round to
13737         odd instructions.
13739 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13741         PR rtl-optimization/77729
13742         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13743         to (X&(C1&~C2))|C2 transformations.
13745 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13747         PR tree-optimization/82363
13748         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13749         mismatch, mark lacc written regardless of racc.
13751 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13753         PR tree-optimization/82381
13754         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13755         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13756         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13757         fallthrough into reassoc_stmt_dominates_stmt_p.
13759         PR target/82386
13760         * combine.c (combine_instructions): Don't combine in unreachable
13761         basic blocks.
13763 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13765         PR target/80210
13766         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13767         function to not use the have_cpu variable.  Do not set cpu_index,
13768         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13769         or the default cpu.
13770         (rs6000_valid_attribute_p): Remove duplicate initializations of
13771         old_optimize and func_optimize.
13772         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13773         (rs6000_activate_target_options): Make global.
13774         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13775         prototype.
13777 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13779         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13780         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13781         Set *poffset to 0 and *psize and *pmax_size to -1 if
13782         *poffset + *psize overflows in HOST_WIDE_INT.
13784         PR tree-optimization/82387
13785         PR tree-optimization/82388
13786         PR tree-optimization/82389
13787         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13788         instead of live_bytes non-NULL.
13790 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13792         PR target/41076
13793         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13794         alternative.
13796 2017-10-02  Richard Biener  <rguenther@suse.de>
13798         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13799         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13800         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13802 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13804         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13805         requested precision matches the type's.
13806         * calls.c (alloc_max_size): Calculate the new candidate size as
13807         a widest_int and use wi::to_widest when comparing it with the
13808         current candidate size.
13809         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13810         zero rather than integer_zero_node.
13811         * match.pd: Check for a no-op conversion before using wi::add
13812         rather than after.  Use tree_to_uhwi when summing small shift
13813         counts into an unsigned int.
13815 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13816             Alan Hayward  <alan.hayward@arm.com>
13817             David Sherwood  <david.sherwood@arm.com>
13819         PR target/71307
13820         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13821         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13822         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13823         POINTER_AND_FP_REGS.
13825 2017-10-02  Richard Biener  <rguenther@suse.de>
13827         PR tree-optimization/82355
13828         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13829         a mapping for the enclosing loop but avoid generating one for
13830         the loop tree root.
13831         (copy_bb_and_scalar_dependences): Remove premature codegen
13832         error on PHIs in blocks duplicated into multiple places.
13833         * graphite-scop-detection.c
13834         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13835         in the region use it as loop and nest to analyze the DR in.
13836         (try_generate_gimple_bb): Likewise.
13837         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13838         (add_loop_constraints): For blocks in a loop not in the region
13839         create a dimension with a single iteration.
13840         * sese.h (gbb_loop_at_index): Remove assert.
13842 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13844         * omp-expand.c (adjust_context_scope): New function.
13845         (expand_parallel_call): Call adjust_context_scope.
13847 2017-10-01  Jeff Law  <law@redhat.com>
13849         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13850         dom_opt_dom_walker class with direct access to private members.
13851         Add comments.  Call test_for_singularity.
13852         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13853         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13854         m_dummy_cond anymore.
13855         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13856         class ctor.
13857         (pass_dominator:execute): Build the dummy_cond here and pass it
13858         to the dom_opt_dom_walker ctor.
13859         (test_for_singularity): New function.
13861 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
13862             Maya Rashish  <coypu@sdf.org>
13864         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
13865         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
13866         (alpha*-*-netbsd*) Use nbsd_tm_file.
13867         (arm*-*-netbsdelf*) Likewise.
13868         (i[34567]86-*-netbsdelf*) Likewise.
13869         (x86_64-*-netbsd*) Likewise.
13870         (mips*-*-netbsd*) Likewise.
13871         (powerpc-*-netbsd*) Likewise.
13872         (sh*-*-netbsd*) Likewise.
13873         (sparc-*-netbsdelf*) Likewise.
13874         (sparc64-*-netbsd*) Likewise.
13875         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
13876         to tm_defines.
13877         (vax-*-netbsdelf*) Likewise.
13878         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
13879         (UINT_FAST8_TYPE) Likewise.
13880         (INT_FAST16_TYPE) Check CHAR_FAST16.
13881         (UINT_FAST16_TYPE) Likewise.
13883 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
13885         PR target/82361
13886         * config/i386/i386.md
13887         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
13888         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
13889         *divmodsi4_zext_2): New define_insn_and_split.
13890         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
13891         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
13892         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
13893         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
13894         New define_insn_and_split.
13895         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
13896         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
13897         operands[1] having DImode when mode is SImode.
13899         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
13900         always SImode for DIV and MOD in REG_EQUAL notes.
13902 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
13904         PR middle-end/82319
13905         * match.pd: Fix handling of NaNs in pattern.
13907 2017-09-29  Jeff Law  <law@redhat.com>
13909         * sbitmap.c (bitmap_bit_in_range_p): New function.
13910         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
13911         * tree-ssa-dse.c (live_bytes_read): New function.
13912         (dse_classify_store): Ignore reads of dead bytes.
13914         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
13915         typos and whitespace errors.
13916         * config/i386/predicates.md (address_no_seg_operand): Likewise.
13917         * config/s390/s390.c (s390_emit_prologue): Likewise.
13919 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13921         PR target/81481
13922         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
13923         with a symbol for LRA.
13925 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13927         PR rtl-optimization/82338
13928         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
13930 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
13932         * genmodes.c (calc_wider_mode): Suppress qsort macro.
13933         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
13934         (qsort_chk): Declare.
13935         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
13936         (qsort_chk): New function.
13938 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13940         PR tree-optimization/82337
13941         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
13942         phi definition if the PHI result appears in an abnormal PHI.
13943         (find_basis_for_base_expr): Don't record a basis if the LHS of the
13944         basis appears in an abnormal PHI.
13946 2017-09-29  Richard Biener  <rguenther@suse.de>
13948         * graphite-isl-ast-to-gimple.c
13949         (translate_isl_ast_to_gimple::set_codegen_error): New function.
13950         (binary_op_to_tree): Use it.
13951         (get_rename_from_scev): Likewise.
13952         (copy_loop_phi_nodes): Likewise.
13953         (copy_bb_and_scalar_dependences): Likewise.
13954         (translate_pending_phi_nodes): Likewise.
13956 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
13958         PR target/82339
13959         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
13960         for movabsq $(i32 << shift), r64.
13962 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
13964         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
13965         index when encoding %esp as %rsp to avoid 0x67 prefix.
13967 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
13969         * config/i386/i386.md (*movsf_internal, *movdf_internal):
13970         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
13972 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
13974         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
13975         Extensions with more than 16 double VFP registers.
13976         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
13977         to_clear_mask and all code related to it.  Replace the remaining
13978         entry by a sbitmap and adapt code accordingly.
13980 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
13982         * brig-builtins.def: Change pure attributes to const.
13984 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
13986         * config.gcc (default_gnu_indirect_function): Default to yes for
13987         sparc*-*-linux* with glibc.
13989 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
13991         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
13992         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
13993         when creating .init_array and .fini_array sections with priority
13994         specified.
13996 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
13998         PR target/71727
13999         * config/aarch64/aarch64.c
14000         (aarch64_builtin_support_vector_misalignment): Always return false
14001         when misalignment is unknown.
14003 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14005         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14006         this function to return false if the definition used by the swap
14007         instruction is artificial, or if the memory address from which the
14008         constant value is loaded is not represented by a base address held
14009         in a register or if the base address register is a frame or stack
14010         pointer.  Additionally, return false if the base address of the
14011         loaded constant is a SYMBOL_REF but is not considered to be a
14012         constant.
14013         (replace_swapped_load_constant): New function.
14014         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14015         loaded constant vector with a load of a swapped constant vector.
14017 2017-09-27  Carl Love  <cel@us.ibm.com>
14019         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14020         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14021         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14022         fctiw instruction.
14024 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14026         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14027         first, always call autopref_rank_data otherwise.
14029 2017-09-27  Richard Biener  <rguenther@suse.de>
14031         * graphite-scop-detection.c (find_scop_parameters): Move
14032         loop bound handling ...
14033         (gather_bbs::before_dom_children): ... here, avoiding the need
14034         to build scop_info->loop_nest.
14035         (record_loop_in_sese): Remove.
14036         * sese.h (sese_info_t::loop_nest): Remove.
14037         * sese.c (new_sese_info): Do not allocate loop_nest.
14038         (free_sese_info): Do not free loop_nest.
14040 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14042         PR c++/82159
14043         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14044         lhs from calls if the lhs has addressable type.
14046 2017-09-27  Richard Biener  <rguenther@suse.de>
14048         * graphite.h (scop::max_alias_set): New member.
14049         * graphite-scop-detection.c: Remove references to non-existing
14050         --param in comments.
14051         (build_alias_sets): Record the maximum alias set used for drs.
14052         (build_scops): Support zero as unlimited for
14053         --param graphite-max-arrays-per-scop.
14054         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14055         and inline into ...
14056         (build_poly_sr_1): ... here.  Compute alias set based on the
14057         maximum alias set used for drs rather than
14058         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14060 2017-09-27  Richard Biener  <rguenther@suse.de>
14062         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14063         --param loop-block-tile-size=0 to disable tiling.
14065 2017-09-27  Richard Biener  <rguenther@suse.de>
14067         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14068         (graphite-max-nb-scop-params): Document special value zero.
14069         * domwalk.h (dom_walker::STOP): New symbolical constant.
14070         (dom_walker::dom_walker): Add optional parameter for bb to
14071         RPO mapping.
14072         (dom_walker::~dom_walker): Declare.
14073         (dom_walker::before_dom_children): Document STOP return value.
14074         (dom_walker::m_user_bb_to_rpo): New member.
14075         (dom_walker::m_bb_to_rpo): Likewise.
14076         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14077         mapping here if not provided by the user.
14078         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14079         provided by the user.
14080         (dom_walker::STOP): Define.
14081         (dom_walker::walk): Do not compute bb to RPO mapping here.
14082         Support STOP return value from before_dom_children to stop
14083         walking.
14084         * graphite-optimize-isl.c (optimize_isl): If the schedule
14085         is the same still generate code if -fgraphite-identity
14086         or -floop-parallelize-all are given.
14087         * graphite-scop-detection.c: Include cfganal.h.
14088         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14089         mapping.
14090         (gather_bbs::before_dom_children): Return STOP for BBs
14091         not in the region.
14092         (build_scops): Compute bb to RPO mapping and pass it to
14093         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14094         as not limiting the number of params.
14095         * graphite.c (graphite_initialize): Remove limit on the
14096         number of basic-blocks in a function.
14097         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14098         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14099         default value of 10.
14101 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14103         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14104         Adjust code to eliminate needing to do the shift right 32-bits
14105         operation after XSCVDPSPN.
14107 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14109         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14110         ((X / Y) != 0 -> X >= Y): Likewise.
14112 2017-09-26  Carl Love  <cel@us.ibm.com>
14114         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14115         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14116         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14117         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14118         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14119         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14120         definitions and overloading.
14121         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14122         statement for P9V_BUILTIN_XST_LEN_R.
14123         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14124         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14125         define_expand and define_insn for the instructions and builtins.
14126         * doc/extend.texi: Update the built-in documentation file for the new
14127         built-in functions.
14128         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14129         define_insn for the instructions
14131 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14133         PR target/39570
14134         * gcc/config/netbsd-protos.h: New file.
14135         * gcc/config/netbsd.c: New file.
14136         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14137         * gcc/config/t-netbsd: New file.
14138         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14139         (tmake_file) Add t-netbsd.
14140         (extra_objs) Add netbsd.o.
14142 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14144         PR fortran/82143
14145         PR fortran/82324
14146         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14148 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14150         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14151         sign extension from a vector register to a GPR by doing a 32-bit
14152         direct move and then an EXTSW.
14153         (extendsi<mode>2 splitter): Likewise.
14154         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14155         right or vector extract after doing XSCVDPSPN.  Use
14156         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14157         the GPRs.
14158         (movdi_from_sf_zero_ext): Likewise.
14159         (reload_gpr_from_vsxsf): Likewise.
14160         (p8_mfvsrd_4_disf): Delete, no longer used.
14161         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14162         then needing to move the SFmode to a GPR to use the XSCVDPSP
14163         instruction instead of FRSP and XSCVDPSPN.
14164         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14165         it is adjacent to the other XSCVSPDP insns.
14166         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14167         SFmode to be in traditional Altivec registers.
14168         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14169         (vsx_xscvspdpn): Likewise.
14170         (vsx_xscvspdpn_scalar): Likewise.
14172 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14174         * tree-sra.c (compare_access_positions): Put integral types first,
14175         stabilize sorting of integral types, remove conditions putting
14176         non-full-precision integers last.
14177         (sort_and_splice_var_accesses): Disable scalarization if a
14178         non-integert would be represented by a non-full-precision integer.
14180 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14182         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14183         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14184         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14185         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14186         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14187         conditionals inside the function instead of around it.  Call
14188         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14189         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14191 2017-09-26  Richard Biener  <rguenther@suse.de>
14193         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14194         fold in ...
14195         (scop_detection::build_scop_breadth): ... this.  Removed.
14196         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14197         (scop_detection::harmful_stmt_in_bb): Likewise.
14198         (scop_detection::graphite_can_represent_stmt): Likewise.
14199         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14200         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14201         (scop_detection::can_represent_loop_1): ... this.  Removed.
14202         (scop_detection::harmful_loop_in_region): Simplify after inlining
14203         the above and remove more quadraticness.
14204         (build_scops): Adjust.
14205         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14206         quadraticness.
14208 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14210         PR target/82267
14211         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14212         REGNO (base) == SP_REG if base is a REG.
14214         PR middle-end/35691
14215         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14216         if it is different SSA_NAME.
14217         (optimize_range_tests_cmp_bitwise): New function.
14218         (optimize_range_tests): Call it.
14220 2017-09-26  Richard Biener  <rguenther@suse.de>
14222         PR tree-optimization/82321
14223         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14224         for the def being inside the loop.
14226 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14228         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14229         assembler output.
14230         * config/s390/s390-builtins.def: Fix constraint on op4.
14232 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14234         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14235         independent expanders.
14236         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14237         ("vec_ordered", "vec_unordered"): New expanders.
14239 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14241         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14242         for SFmode.
14244 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14246         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14247         vec_unpacks_lo_v16qi.
14248         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14250 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14252         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14253         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14254         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14256 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14258         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14259         predicate.
14260         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14261         and V16QI.
14262         ("*vec_slb<mode>"): New insn pattern.
14263         ("vec_shr_<mode>"): New expander.
14264         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14265         and force the shift count operand to V16QImode.
14266         ("vec_srb<mode>"): Set shift count mode to V16QI.
14268 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14270         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14271         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14272         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14274 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14276         PR target/82175
14277         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14279 2017-09-26  Richard Biener  <rguenther@suse.de>
14281         PR tree-optimization/82320
14282         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14283         isn't a change.
14285 2017-09-25  Jeff Law  <law@redhat.com>
14287         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14288         prototype for new argument.
14289         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14290         mostly extracted from rs6000_emit_allocate_stack.
14291         (rs6000_emit_probe_stack_range_stack_clash): New function.
14292         (rs6000_emit_allocate_stack): Call
14293         rs6000_emit_probe_stack_range_stack_clash as needed.
14294         (rs6000_emit_probe_stack_range): Add additional argument
14295         to call to gen_probe_stack_range{si,di}.
14296         (output_probe_stack_range): New.
14297         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14298         (output_probe_stack_range_stack_clash): New.
14299         (rs6000_emit_prologue): Emit notes into dump file as requested.
14300         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14301         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14302         Add additional operand and pass it to output_probe_stack_range.
14304 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14306         PR tree-optimization/82163
14307         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14308         (checking_verify_loop_closed_ssa): New parameter.
14309         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14310         (check_loop_closed_ssa_stmt): Delete.
14311         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14312         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14313         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14314         changed loops.
14316 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14318         * brig-builtins.def: Treat HSAIL barrier builtins as
14319         setjmp/longjump style functions.
14321 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14323         * target.def (constant_alignment): New hook.
14324         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14325         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14326         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14327         * doc/tm.texi: Regenerate.
14328         * targhooks.h (default_constant_alignment): Declare.
14329         (constant_alignment_word_strings): Likewise.
14330         * targhooks.c (default_constant_alignment): New function.
14331         (constant_alignment_word_strings): Likewise.
14332         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14333         instead of CONSTANT_ALIGNMENT.
14334         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14335         (force_const_mem): Likewise.
14336         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14337         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14338         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14339         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14340         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14341         definition.
14342         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14343         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14344         constant_alignment_word_strings.
14345         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14346         (CONSTANT_ALIGNMENT): Likewise.
14347         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14348         (arm_constant_alignment): New function.
14349         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14350         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14351         constant_alignment_word_strings.
14352         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14353         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14354         constant_alignment_word_strings.
14355         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14356         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14357         (cris_constant_alignment): New function.
14358         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14359         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14360         (epiphany_constant_alignment): New function.
14361         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14362         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14363         constant_alignment_word_strings.
14364         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14365         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14366         constant_alignment_word_strings.
14367         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14368         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14369         constant_alignment_word_strings.
14370         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14371         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14372         * config/i386/i386.c (ix86_constant_alignment): Make static.
14373         Use the same interface as the target hook.
14374         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14375         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14376         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14377         constant_alignment_word_strings.
14378         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14379         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14380         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14381         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14382         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14383         constant_alignment_word_strings.
14384         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14385         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14386         constant_alignment_word_strings.
14387         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14388         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14389         constant_alignment_word_strings.
14390         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14391         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14392         New function.
14393         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14394         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14395         * config/mips/mips.c (mips_constant_alignment): New function.
14396         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14397         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14398         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14399         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14400         (mmix_constant_alignment): Make static.  Use the same interface
14401         as the target hook.
14402         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14403         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14404         constant_alignment_word_strings.
14405         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14406         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14407         constant_alignment_word_strings.
14408         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14409         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14410         constant_alignment_word_strings.
14411         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14412         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14413         (rs6000_constant_alignment): New function.
14414         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14415         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14416         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14417         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14418         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14419         (rs6000_constant_alignment): New function.
14420         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14421         * config/s390/s390.c (s390_constant_alignment): New function.
14422         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14423         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14424         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14425         constant_alignment_word_strings.
14426         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14427         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14428         (sparc_constant_alignment): New function.
14429         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14430         * config/spu/spu.c (spu_constant_alignment): New function.
14431         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14432         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14433         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14434         constant_alignment_word_strings.
14435         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14436         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14437         constant_alignment_word_strings.
14438         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14439         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14440         constant_alignment_word_strings.
14441         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14442         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14443         (visium_constant_alignment): New function.
14444         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14445         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14446         (xtensa_constant_alignment): New function.
14447         * system.h (CONSTANT_ALIGNMENT): Poison.
14449 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14451         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14452         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14453         (rs6000_builtin_valid_without_lhs): New helper function.
14454         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14455         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14457 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14459         * target.h (vec_perm_indices): Use unsigned short rather than
14460         unsigned char.
14461         (auto_vec_perm_indices): Likewise.
14462         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14463         Use unsigned int rather than unsigned char.
14464         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14466 2017-09-25  Richard Biener  <rguenther@suse.de>
14468         * cfgloop.h (sort_sibling_loops): Declare.
14469         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14470         (sort_sibling_loops): New function sorting the sibling loop list
14471         in RPO order.
14472         * graphite.c (graphite_transform_loops): Sort sibling loops.
14474 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14476         * target.def (vec_perm_const_ok): Change sel parameter to
14477         vec_perm_indices.
14478         * optabs-query.c (can_vec_perm_p): Update accordingly.
14479         * doc/tm.texi: Regenerate.
14480         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14481         auto_vec_perm_indices and remove separate nelt field.
14482         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14483         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14484         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14485         (aarch64_expand_vec_perm_const): Update accordingly.
14486         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14487         to vec_perm_indices.
14488         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14489         auto_vec_perm_indices and remove separate nelt field.
14490         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14491         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14492         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14493         accordingly.
14494         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14495         to vec_perm_indices.
14496         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14497         sel to vec_perm_indices.
14498         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14499         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14500         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14501         Likewise.
14502         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14503         Likewise.
14505 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14507         PR debug/82155
14508         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14509         on the FUNCTION_DECL function context if it has a DIE that is a
14510         declaration.
14512 2017-09-25  Richard Biener  <rguenther@suse.de>
14514         PR tree-optimization/82285
14515         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14516         enumeral types.
14518 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14520         PR target/80035
14521         PR target/81069
14522         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14523         noreturn function.
14525 2017-09-25  Richard Biener  <rguenther@suse.de>
14527         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14528         ISL errors other than isl_error_quota happen.  Dump if the
14529         schedule is the same.
14530         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14531         errors instead of aborting inside ISL.
14533 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14535         PR target/80556
14536         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14537         of libgcc_eh for m64.
14538         * config/i386/darwin64.h: Likewise.
14540 2017-09-25  Richard Biener  <rguenther@suse.de>
14542         PR middle-end/82144
14543         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14544         attribute for incomplete types nor twice for complete ones.
14546 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14548         PR target/82267
14549         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14550         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14551         register.
14553 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14555         PR bootstrap/82306
14556         * config/i386/i386.opt (mprefer-avx256): Use
14557         ix86_target_flags variable.
14558         * config/i386/i386.c (ix86_target_string): Move
14559         -mprefer-avx256 to flag2_opts.
14561 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14563         PR middle-end/35691
14564         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14565         and x != -1 | y != -1 into (x & y) != -1.
14567 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14569         * config.gcc: Add new case statement to set
14570         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14571         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14572         s390x-*-linux* case statements.   Added aarch64 to the list of
14573         supported architectures.
14575 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14577         PR tree-optimization/82289
14578         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14579         STMT_VINFO_RELEVANT_P.
14581 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14582             Alan Hayward  <alan.hayward@arm.com>
14583             David Sherwood  <david.sherwood@arm.com>
14585         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14586         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14587         (extract_range_from_binary_expr_1): Don't call
14588         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14590 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14591             Alan Hayward  <alan.hayward@arm.com>
14592             David Sherwood  <david.sherwood@arm.com>
14594         * target.def (preferred_vector_alignment): New hook.
14595         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14596         hook.
14597         * doc/tm.texi: Regenerate.
14598         * targhooks.h (default_preferred_vector_alignment): Declare.
14599         * targhooks.c (default_preferred_vector_alignment): New function.
14600         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14601         Expand commentary.
14602         (DR_TARGET_ALIGNMENT): New macro.
14603         (aligned_access_p): Update commentary.
14604         (vect_known_alignment_in_bytes): New function.
14605         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14606         function.
14607         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14608         Calculate the misalignment based on the target alignment rather than
14609         the vector size.
14610         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14611         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14612         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14613         the target alignment, rather than masking the element misalignment
14614         with the number of elements in a vector.  Also use the target
14615         alignment when calculating the maximum number of peels.
14616         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14617         instead of TYPE_ALIGN_UNIT.
14618         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14619         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14620         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14621         (vect_create_data_ref_ptr): Likewise.
14622         (vect_setup_realignment): Realign by ANDing with
14623         -DR_TARGET_MISALIGNMENT.
14624         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14625         the number of peels based on DR_TARGET_ALIGNMENT.
14626         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14627         with the guaranteed alignment boundary when deciding whether
14628         overrun is OK.
14629         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14630         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14631         (ensure_base_align): Remove stmt_info parameter.  Get the
14632         target base alignment from DR_TARGET_ALIGNMENT.
14633         (vectorizable_store): Update call accordingly.   Interpret
14634         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14635         TYPE_ALIGN_UNIT.
14636         (vectorizable_load): Likewise.
14638 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14639             Alan Hayward  <alan.hayward@arm.com>
14640             David Sherwood  <david.sherwood@arm.com>
14642         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14643         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14644         (vect_enhance_data_refs_alignment): Likewise.
14646 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14648         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14649         error.  Only quit immediately if parsing is complete.
14650         (BEGIN): Initialize fatal_err and parse_done.
14651         (begin fpu, end fpu): Check number of arguments.
14652         (begin arch, end arch): Likewise.
14653         (begin cpu, end cpu): Likewise.
14654         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14655         (optalias): Likewise.
14657 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14659         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14660         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14661         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14662         * config/arm/arm.c (arm_option_override): Use lower case for feature
14663         bit names.
14664         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14665         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14666         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14667         (isa_pfx): Delete.
14668         (print_isa_bits_for): New function.
14669         (gen_isa): New function.
14670         (gen_comm_data): Use print_isa_bits_for.
14671         (define feature): New keyword.
14672         (define fgroup): New keyword.
14673         * config/arm/t-arm (TM_H): Remove.
14674         (GTM_H): Add arm-isa.h.
14675         (arm-isa.h): Add rule to generate file.
14676         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14677         case for feature bit names.
14679 2017-09-22  Richard Biener  <rguenther@suse.de>
14681         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14682         single caller.
14683         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14684         print of no dependency loops ...
14685         * graphite.c (graphite_transform_loops): ... here.
14686         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14687         loops.
14688         (same_close_phi_node, remove_duplicate_close_phi,
14689         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14690         (canonicalize_loop_closed_ssa): ... here and simplify.
14691         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14692         (optimize_isl): Use dump_printf_loc to tell when we stopped
14693         optimizing because of an ISL timeout.
14695 2017-09-22  Richard Biener  <rguenther@suse.de>
14697         PR tree-optimization/82291
14698         * tree-if-conv.c (predicate_mem_writes): Make sure to
14699         remove writes in blocks predicated with false.
14701 2017-09-22  Richard Biener  <rguenther@suse.de>
14703         * sese.c: Include cfganal.h.
14704         (if_region_set_false_region): Remove.
14705         (create_if_region_on_edge): Likewise.
14706         (move_sese_in_condition): Re-implement without destroying
14707         dominators.
14709 2017-09-22  Richard Biener  <rguenther@suse.de>
14711         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14712         Verify both BBs contain loop PHI nodes before dispatching to
14713         copy_loop_phi_args.
14714         (graphite_regenerate_ast_isl): Do not recompute dominators,
14715         do not verify three times.  Restructure for clarity.
14716         * graphite-scop-detection.c (same_close_phi_node,
14717         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14718         defined_in_loop_p, canonicalize_loop_closed_ssa,
14719         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14720         checking and SSA rewrite, move to ...
14721         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14722         (graphite_initialize): Do not pass in ctx, do not reset the
14723         SCEV cache, compute only dominators.
14724         (graphite_transform_loops): Allocate ISL ctx after
14725         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14726         Maintain post-dominators only around build_scops.
14727         * sese.c (if_region_set_false_region): Make static.  Free
14728         and recompute dominators.
14729         (move_sese_in_condition): Assert we don't get called with
14730         post-dominators computed.
14731         * sese.h (if_region_set_false_region): Remove.
14733 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14735         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14736         mode attribute for TARGET_AVX512VL.
14738 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14740         * config/i386/i386.opt (mprefer-avx256): New option.
14741         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14742         to flag_opts.
14743         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14744         for TARGET_PREFER_AVX256.
14745         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14747 2017-09-21  Jeff Law  <law@redhat.com>
14749         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14750         Fix dump output if the only stack space is for pushed registers.
14752 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14754         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14755         of insn_cost.
14757 2017-09-21  Martin Sebor  <msebor@redhat.com>
14759         PR c/81882
14760         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14761         code (in C++) or code that triggers warnings.
14763 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14765         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14767 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14769         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14770         * sched-rgn.c: Ditto.
14771         * sel-sched-ir.c: Ditto.
14773 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14775         * toplev.h (set_random_seed): Adjust return type.
14776         * toplev.c (init_local_tick): Move eager initialization of random_seed
14777         to get_random_seed.  Adjust comment.
14778         (init_random_seed): Inline to get_random_seed, delete.
14779         (get_random_seed): Initialize random_seed lazily.
14780         (set_random_seed): Do not return previous value.
14781         (print_switch_value): Do not call get_random_seed.
14783 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14785         * cgraph.c (delete_function_version): New, broken out from...
14786         (cgraph_node::delete_function_version): ...here.  Rename to
14787         cgraph_node::delete_function_version_by_decl.  Update all uses.
14788         (cgraph_node::remove): Call delete_function_version.
14790 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14792         PR sanitizer/81715
14793         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14794         VAR_DECLs to which addressable non-volatile parameters are mapped
14795         and for id->retvar after the return value assignment.  Clear
14796         id->retval and id->retbnd after inlining.
14798 2017-09-21  Richard Biener  <rguenther@suse.de>
14800         PR tree-optimization/82276
14801         PR tree-optimization/82244
14802         * tree-vrp.c (build_assert_expr_for): Set
14803         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14804         has it set.
14805         (remove_range_assertions): Revert earlier change.
14807 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14809         PR target/71951
14810         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14812 2017-09-21  Richard Biener  <rguenther@suse.de>
14814         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14815         Restore valid IL after code generation errors.
14816         * graphite.c (graphite_transform_loops): Diagnose code
14817         generation issues as MSG_MISSED_OPTIMIZATION and continue
14818         with processing SCOPs.
14820 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14821             Alan Hayward  <alan.hayward@arm.com>
14822             David Sherwood  <david.sherwood@arm.com>
14824         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14825         rather than choosing between plus_constant and gen_rtx_<CODE>.
14826         * expr.c (emit_push_insn): Likewise.
14827         (expand_expr_real_2): Likewise.
14829 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14830             Alan Hayward  <alan.hayward@arm.com>
14831             David Sherwood  <david.sherwood@arm.com>
14833         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14835 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14836             Alan Hayward  <alan.hayward@arm.com>
14837             David Sherwood  <david.sherwood@arm.com>
14839         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14840         calling tree_to_uhwi.
14842 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14843             Alan Hayward  <alan.hayward@arm.com>
14844             David Sherwood  <david.sherwood@arm.com>
14846         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14847         INTEGER_CST rather than a negative test for ADDR_EXPR.
14849 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14850             Alan Hayward  <alan.hayward@arm.com>
14851             David Sherwood  <david.sherwood@arm.com>
14853         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14854         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14856 2017-09-21  Richard Biener  <rguenther@suse.de>
14858         PR tree-optimization/71351
14859         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14860         graphite_create_new_loop_guard): Remove, fold remaining parts
14861         into caller ...
14862         (translate_isl_ast_node_for): ... here and simplify.
14864 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14866         PR target/82260
14867         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
14868         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
14869         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
14870         alternative always use QI mode, for -Os imov (=R,R) alternative
14871         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
14872         ignore -Os.
14874 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14875             Jeff Law  <law@redhat.com>
14877         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
14878         (allocate_stack_space): New function, partially extracted from
14879         s390_emit_prologue.
14880         (s390_emit_prologue): Track offset to most recent stack probe.
14881         Code to allocate space moved into allocate_stack_space.
14882         Dump actions when no stack is allocated.
14883         (s390_prologue_plus_offset): New function.
14884         (s390_emit_stack_probe): Likewise.
14886 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
14888         * common.opt (Wa, Wl, Wp, g, gz=): Add
14889         RejectNegative.
14890         (gno-column-info): Remove.
14891         (gcolumn-info): Drop RejectNegative.
14892         (gno-): New prefix.
14893         (gno-record-gcc-switches): Remove.
14894         (grecord-gcc-switches): Drop RejectNegative.
14895         (gno-split-dwarf): Remove.
14896         (gsplit-dwarf): Drop RejectNegative.
14897         (gno-strict-dwarf): Remove.
14898         (gstrict-dwarf): Drop RejectNegative.
14899         * config/darwin.opt (gfull, gused): Add RejectNegative.
14900         * dwarf2out.c (gen_producer_string): Drop
14901         gno-record-gcc-switches handler.
14902         * optc-gen.awk: Add g to prefixes with negative forms.
14903         * opts-common.c (remapping_prefix_p): New.
14904         (find_opt): Check it.
14905         (generate_canonical_option): Test g prefix.
14906         (option_map): Add -gno- mapping.
14907         (add_misspelling_candidates): Check remapping_prefix_p.
14909 2017-09-20  Jeff Law  <law@redhat.com>
14911         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
14912         thinko in stack clash protection support.
14914         * explow.c (compute_stack_clash_protection_loop_data): Use
14915         CONST_INT_P instead of explicit test.  Verify object is a
14916         CONST_INT_P before looking at INTVAL.
14917         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
14918         instead of explicit test.
14920 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
14922         PR target/77687
14923         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
14924         address instead of to r1 and r11.
14926 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
14928         * config.gcc: Support "knm".
14929         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
14930         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14931         PROCESSOR_KNM.
14932         * config/i386/i386.c (m_KNM): Define.
14933         (processor_target_table): Add "knm".
14934         (PTA_KNM): Define.
14935         (ix86_option_override_internal): Add "knm".
14936         (ix86_issue_rate): Add PROCESSOR_KNM.
14937         (ix86_adjust_cost): Ditto.
14938         (ia32_multipass_dfa_lookahead): Ditto.
14939         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
14940         (fold_builtin_cpu): Add M_INTEL_KNM.
14941         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
14942         (processor_type): Add PROCESSOR_KNM.
14943         * config/i386/x86-tune.def: Add m_KNM.
14944         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
14946 2017-09-20  Richard Biener  <rguenther@suse.de>
14948         PR tree-optimization/80213
14949         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
14950         are allowed in empty BBs as well.
14951         (canonicalize_loop_closed_ssa): Also look for other complex
14952         edges.
14953         (scop_detection::get_sese): Include the loop-closed PHI block
14954         in loop SESEs.
14955         (scop_detection::merge_sese): Remove code adding extra blocks.
14956         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
14957         (build_scops): Assert the final returned scop is invalid.
14959 2017-09-20  Richard Biener  <rguenther@suse.de>
14961         PR tree-optimization/82264
14962         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
14963         for GIMPLE_CONDs.
14964         (vn_phi_lookup): Likewise.
14965         (vn_phi_insert): Likewise.
14967 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
14969         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
14970         that fits into uhwi or shwi, add DW_AT_const_value regardless
14971         of early_dwarf without going through RTL, using add_AT_unsigned
14972         or add_AT_int.
14974         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
14975         (DEBUG_LTO_ABBREV_SECTION): Likewise.
14976         (DEBUG_LTO_MACINFO_SECTION): Likewise.
14977         (DEBUG_MACRO_SECTION): Likewise.
14978         (DEBUG_LTO_MACRO_SECTION): Likewise.
14979         (DEBUG_STR_DWO_SECTION): Likewise.
14980         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
14981         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
14982         (DEBUG_LTO_DWO_LINE_SECTION): Define.
14983         (DEBUG_LTO_LINE_STR_SECTION): Define.
14984         (init_sections_and_labels): Initialize debug_line_str_section
14985         variable.  Initialize debug_loc_section for -gdwarf-5 to
14986         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
14988 2017-09-20  Richard Biener  <rguenther@suse.de>
14990         * graphite-sese-to-poly.c (extract_affine): Properly handle
14991         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
14993 2017-09-20  Richard Biener  <rguenther@suse.de>
14995         PR tree-optimization/81373
14996         * graphite-scop-detection.c (build_cross_bb_scalars_def):
14997         Force SESE live-out defs to be handled even if they are
14998         scev_analyzable_p.
15000 2017-09-19  Jeff Law  <law@redhat.com>
15002         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15003         nothing for stack adjustments with REG_STACK_CHECK.
15004         * sched-deps.c (parse_add_or_inc): Reject insns with
15005         REG_STACK_CHECK from dependency breaking.
15006         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15007         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15008         * reg-notes.def (STACK_CHECK): New note.
15010         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15011         (ix86_expand_prologue): Dump stack clash info as needed.
15012         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15014         * function.c (dump_stack_clash_frame_info): New function.
15015         * function.h (dump_stack_clash_frame_info): Prototype.
15016         (enum stack_clash_probes): New enum.
15018         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15019         flag_stack_clash_protection.
15020         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15021         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15022         (arm_frame_pointer_required): Likewise.
15023         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15024         (ia64_expand_prologue): Likewise.
15025         * config/mips/mips.c (mips_expand_prologue): Likewise.
15026         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15027         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15028         (sparc_flat_expand_prologue): Likewise.
15029         * config/spu/spu.c (spu_expand_prologue): Likewise.
15031         * explow.c: Include "params.h".
15032         (anti_adjust_stack_and_probe_stack_clash): New function.
15033         (get_stack_check_protect): Likewise.
15034         (compute_stack_clash_protection_loop_data): Likewise.
15035         (emit_stack_clash_protection_loop_start): Likewise.
15036         (emit_stack_clash_protection_loop_end): Likewise.
15037         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15038         Use anti_adjust_stack_and_probe_stack_clash.
15039         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15040         (emit_stack_clash_protection_loop_start): Likewise.
15041         (emit_stack_clash_protection_loop_end): Likewise.
15042         * rtl.h (get_stack_check_protect): Prototype.
15043         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15044         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15045         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15046         Prototype.
15047         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15048         Add @hook.
15049         * doc/tm.texi: Rebuilt.
15050         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15051         get_stack_check_protect.
15052         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15053         * config/arm/arm.c (arm_expand_prologue): Likewise.
15054         (arm_frame_pointer_required): Likewise.
15055         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15056         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15057         * config/mips/mips.c (mips_expand_prologue): Likewise.
15058         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15059         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15060         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15061         (sparc_flat_expand_prologue): Likewise.
15063         * common.opt (-fstack-clash-protection): New option.
15064         * flag-types.h (enum stack_check_type): Note difference between
15065         -fstack-check= and -fstack-clash-protection.
15066         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15067         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15068         * toplev.c (process_options): Issue warnings/errors for cases
15069         not handled with -fstack-clash-protection.
15070         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15071         (-fstack-check): Note additional problem with -fstack-check=generic.
15072         Note that -fstack-check is primarily for Ada and refer users
15073         to -fstack-clash-protection for stack-clash-protection.
15074         Document new params for stack clash protection.
15076 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15078         * config/i386/i386.c (ix86_split_long_move): Do not handle
15079         address used for LEA in a special way.
15081 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15083         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15085 2017-09-19  Martin Sebor  <msebor@redhat.com>
15087         PR c/81854
15088         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15089         of incompatible types.
15091 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15093         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15094         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15095         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15096         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15098 2017-09-19  Richard Biener  <rguenther@suse.de>
15100         PR tree-optimization/82244
15101         * tree-vrp.c (remove_range_assertions): Do not propagate
15102         a constant to abnormals but replace the assert with a copy.
15104 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15106         PR rtl-optimization/57878
15107         PR rtl-optimization/68988
15108         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15109         avoidance test involving non_reload_pseudos.  Move frequency test
15110         below the general fragmentation avoidance test.
15112 2017-09-19  Richard Biener  <rguenther@suse.de>
15114         PR tree-optimization/69728
15115         * graphite-sese-to-poly.c (schedule_error): New global.
15116         (add_loop_schedule): Handle empty domain by failing the
15117         schedule.
15118         (build_original_schedule): Handle schedule_error.
15120 2017-09-19  Richard Biener  <rguenther@suse.de>
15122         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15123         Do not iterate to sibling loops but only to siblings of inner
15124         loops.
15126 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15128         PR target/81613
15129         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15130         that the registers are different.
15132 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15134         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15135         to processor_model and "amdfam17h" to arch_names_table.
15136         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15138 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15140         PR c/82234
15141         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15143 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15144             Alan Hayward  <alan.hayward@arm.com>
15145             David Sherwood  <david.sherwood@arm.com>
15147         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15148         with a vec_info *.
15149         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15150         accordingly.
15151         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15152         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15153         vect_schedule_slp_instance.
15154         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15155         Update call to vect_slp_analyze_node_operations.  Simplify return
15156         value.
15157         (vect_slp_analyze_bb_1): Update call accordingly.
15158         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15159         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15160         (vect_schedule_slp): Update call accordingly.
15162 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15163             Alan Hayward  <alan.hayward@arm.com>
15164             David Sherwood  <david.sherwood@arm.com>
15166         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15167         with types that aren't in fact scalar.
15169 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15171         * tree-vect-slp.c (vect_record_max_nunits): New function,
15172         split out from...
15173         (vect_build_slp_tree_1): ...here.
15174         (vect_build_slp_tree_2): Call it for phis too.
15176 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15178         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15179         to vect_get_vec_def_for_operand when getting the mask operand.
15181 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15182             Alan Hayward  <alan.hayward@arm.com>
15183             David Sherwood  <david.sherwood@arm.com>
15185         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15186         bitstart.
15188 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15189             Alan Hayward  <alan.hayward@arm.com>
15190             David Sherwood  <david.sherwood@arm.com>
15192         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15193         calculation for vector booleans.
15195 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15196             Alan Hayward  <alan.hayward@arm.com>
15197             David Sherwood  <david.sherwood@arm.com>
15199         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15200         split out from...
15201         (vect_transform_stmt): ...here.
15202         (vect_analyze_stmt): Use it instead of calling
15203         vectorizable_live_operation directly.
15205 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15207         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15208         non-SIMT targets in acc vector loops.
15210 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15212         * configure.ac: Add arc and check if assembler supports gdwarf2.
15213         * configure: Regenerate.
15215 2017-09-18  Richard Biener  <rguenther@suse.de>
15217         PR tree-optimization/82220
15218         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15219         epilogue niters from the min_profitable_iters compute.
15221 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15223         PR target/82145
15224         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15225         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15226         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15228 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15230         PR target/81361
15231         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15232         switching to a new text section.
15234 2017-09-18  Richard Biener  <rguenther@suse.de>
15236         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15237         Simplify.
15238         (build_alias_set): Reject aliases with no access function.
15240 2017-09-18  Richard Biener  <rguenther@suse.de>
15242         PR tree-optimization/79622
15243         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15244         handle PHIs.
15245         (build_cross_bb_scalars_use): Likewise.
15247 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15249         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15251 2017-09-18  Alan Modra  <amodra@gmail.com>
15253         PR target/81996
15254         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15255         stack_pointer_rtx for count 0.  Update comments.  Break up
15256         large rtl expression.
15258 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15260         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15261         Increase to 20 bytes.
15262         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15263         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15264         or avx version of the stub.
15266 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15268         PR target/82166
15269         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15270         compute the minimum stack alignment.  Also update preferred stack
15271         boundary for leaf functions.
15273 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15275         PR tree-optimization/82228
15276         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15277         of ncopies.
15279 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15281         * common/config/nds32/nds32-common.c
15282         (nds32_option_optimization_table): Refine formatting.
15283         (nds32_option_optimization_table): Use -fsched-pressure and
15284         -fomit-frame-pointer for specific optimization level.
15286 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15288         * config/nds32/nds32.c: Refine formatting and comments.
15289         * config/nds32/nds32.h: Likewise.
15290         * config/nds32/nds32.md: Likewise.
15291         * config/nds32/nds32-cost.c: Likewise.
15292         * config/nds32/nds32-isr.c: Likewise.
15293         * config/nds32/nds32-md-auxiliary.c: Likewise.
15294         * config/nds32/nds32-multiple.md: Likewise.
15295         * config/nds32/nds32-predicates.c: Likewise.
15297 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15298             Jakub Jelinek  <jakub@redhat.com>
15300         Add support for -std=c++2a.
15301         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15302         or -std=gnu+2a.
15303         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15305 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15307         PR target/82066
15308         * doc/extend.texi (Common Function Attributes): Add 
15309         references to ARM, AArch64, and S/390 specific attributes.
15310         (Function Specific Option Pragmas): Add AArch64 and S/390
15311         to list of back ends that support the target pragma.
15313 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15315         * doc/standards.texi: Fix C++17 description.  Update URLs for
15316         C++11 & 14.
15318 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15320         * common.opt (Wcast-align=strict): New warning option.
15321         * doc/invoke.texi: Document -Wcast-align=strict. 
15323 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15325         * cgraph.h (cgraph_thunk_info): Add comments.
15326         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15327         assert for VIRTUAL_* arguments stricter.
15329 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15331         PR tree-optimization/71026
15332         * match.pd: Move RDIV patterns from fold-const.c
15333         * fold-const.c (distribute_real_division): Removed.
15334         (fold_binary_loc): Remove calls to distribute_real_divison.
15336 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15338         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15339         c++1z and gnu++1z as deprecated.  Change other references to
15340         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15341         Change -Wc++1z-compat to -Wc++17-compat.
15342         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15343         * dwarf2out.c (highest_c_language): Handle C++17.
15344         (gen_compile_unit_die): Likewise.
15346 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15348         PR rtl-optimization/82192
15349         * combine.c (make_extraction): Don't look through non-paradoxical
15350         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15351         inner's mode.
15353 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15354             Alan Hayward  <alan.hayward@arm.com>
15355             David Sherwood  <david.sherwood@arm.com>
15357         * target.def (function_arg_offset): New hook.
15358         * targhooks.h (default_function_arg_offset): Declare.
15359         * targhooks.c (default_function_arg_offset): New function.
15360         * function.c (locate_and_pad_parm): Use
15361         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15362         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15363         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15364         * doc/tm.texi: Regenerate.
15365         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15366         * config/spu/spu.c (spu_function_arg_offset): New function.
15367         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15368         * system.h (FUNCTION_ARG_OFFSET): Poison.
15370 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15371             Alan Hayard  <alan.hayward@arm.com>
15372             David Sherwood  <david.sherwood@arm.com>
15374         * target.def (truly_noop_truncation): New hook.
15375         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15376         than TRULY_NOOP_TRUNCATION.
15377         * hooks.h (hook_bool_uint_uint_true): Declare.
15378         * hooks.c (hook_bool_uint_uint_true): New function.
15379         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15380         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15381         * doc/tm.texi: Regenerate.
15382         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15383         rather than TRULY_NOOP_TRUNCATION in comments.
15384         (simplify_comparison): Likewise.
15385         (record_truncated_value): Likewise.
15386         * expmed.c (extract_bit_field_1): Likewise.
15387         (extract_split_bit_field): Likewise.
15388         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15389         instead of TRULY_NOOP_TRUNCATION.
15390         * function.c (assign_parm_setup_block): Likewise.
15391         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15392         * rtlhooks.c: Include target.h.
15393         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15394         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15395         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15396         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15397         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15398         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15399         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15400         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15401         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15402         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15403         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15404         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15405         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15406         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15407         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15408         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15409         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15410         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15411         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15412         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15413         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15414         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15415         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15416         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15417         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15418         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15419         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15420         rather than TRULY_NOOP_TRUNCATION in comments.
15421         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15422         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15423         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15424         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15425         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15426         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15427         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15428         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15429         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15430         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15431         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15432         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15433         rather than TRULY_NOOP_TRUNCATION in comments.
15434         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15435         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15436         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15437         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15438         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15439         TRULY_NOOP_TRUNCATION condition.
15440         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15441         (TRULY_NOOP_TRUNCATION): Delete.
15442         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15443         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15444         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15445         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15446         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15447         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15448         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15449         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15450         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15451         rather than TRULY_NOOP_TRUNCATION in comments.
15452         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15453         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15454         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15455         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15456         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15457         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15459 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15461         PR target/67591
15462         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15463         (*cmp_ior): Likewise.
15464         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15465         (*ior_scc_scc_cmp): Likewise.
15466         (*and_scc_scc): Likewise.
15467         (*and_scc_scc_cmp): Likewise.
15469 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15470             Alan Hayard  <alan.hayward@arm.com>
15471             David Sherwood  <david.sherwood@arm.com>
15473         * target.def (can_change_mode_class): New hook.
15474         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15475         (hard_regno_nregs): Likewise.
15476         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15477         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15478         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15479         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15480         (LOAD_EXTEND_OP): Update accordingly.
15481         * doc/tm.texi: Regenerate.
15482         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15483         CANNOT_CHANGE_MODE_CLASS.
15484         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15485         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15486         * combine.c (simplify_set): Update accordingly.
15487         * emit-rtl.c (validate_subreg): Likewise.
15488         * recog.c (general_operand): Likewise.
15489         * regcprop.c (mode_change_ok): Likewise.
15490         * reload1.c (choose_reload_regs): Likewise.
15491         (inherit_piecemeal_p): Likewise.
15492         * rtlanal.c (simplify_subreg_regno): Likewise.
15493         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15494         instead of CANNOT_CHANGE_MODE_CLASS.
15495         (reload_cse_simplify_operands): Likewise.
15496         * reload.c (push_reload): Use targetm.can_change_mode_class
15497         instead of CANNOT_CHANGE_MODE_CLASS.
15498         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15499         REG_CANNOT_CHANGE_MODE_P.
15500         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15501         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15502         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15503         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15504         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15505         (arm_can_change_mode_class): New function.
15506         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15507         than CANNOT_CHANGE_MODE_CLASS in comments.
15508         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15509         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15510         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15511         (ix86_can_change_mode_class): ...this new function, inverting the
15512         sense of the return value.
15513         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15514         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15515         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15516         (ia64_can_change_mode_class): New function.
15517         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15518         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15519         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15520         (m32c_can_change_mode_class): ...this new function, inverting the
15521         sense of the return value.
15522         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15523         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15524         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15525         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15526         (mips_can_change_mode_class): ...this new function, inverting the
15527         sense of the return value.
15528         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15529         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15530         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15531         (msp430_can_change_mode_class): New function.
15532         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15533         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15534         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15535         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15536         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15537         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15538         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15539         (pa_cannot_change_mode_class): Replace with...
15540         (pa_can_change_mode_class): ...this new function, inverting the
15541         sense of the return value.
15542         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15543         than CANNOT_CHANGE_MODE_CLASS in comments.
15544         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15545         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15546         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15547         (pdp11_cannot_change_mode_class): Replace with...
15548         (pdp11_can_change_mode_class): ...this new function, inverting the
15549         sense of the return value.
15550         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15551         * config/powerpcspe/powerpcspe-protos.h
15552         (rs6000_cannot_change_mode_class_ptr): Delete.
15553         * config/powerpcspe/powerpcspe.c
15554         (rs6000_cannot_change_mode_class_ptr): Delete.
15555         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15556         (rs6000_option_override_internal): Assign to
15557         targetm.can_change_mode_class instead of
15558         rs6000_cannot_change_mode_class_ptr.
15559         (rs6000_cannot_change_mode_class): Replace with...
15560         (rs6000_can_change_mode_class): ...this new function, inverting the
15561         sense of the return value.
15562         (rs6000_debug_cannot_change_mode_class): Replace with...
15563         (rs6000_debug_can_change_mode_class): ...this new function.
15564         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15565         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15566         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15567         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15568         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15569         Delete.
15570         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15571         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15572         (rs6000_option_override_internal): Assign to
15573         targetm.can_change_mode_class instead of
15574         rs6000_cannot_change_mode_class_ptr.
15575         (rs6000_cannot_change_mode_class): Replace with...
15576         (rs6000_can_change_mode_class): ...this new function, inverting the
15577         sense of the return value.
15578         (rs6000_debug_cannot_change_mode_class): Replace with...
15579         (rs6000_debug_can_change_mode_class): ...this new function.
15580         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15581         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15582         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15583         (s390_can_change_mode_class): ...this new function, inverting the
15584         sense of the return value.
15585         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15586         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15587         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15588         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15589         (sh_cannot_change_mode_class): Replace with...
15590         (sh_can_change_mode_class): ...this new function, inverting the
15591         sense of the return value.
15592         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15593         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15594         (sparc_can_change_mode_class): New function.
15595         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15596         * config/spu/spu.c (spu_can_change_mode_class): New function.
15597         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15598         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15599         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15600         (visium_can_change_mode_class): New function.
15601         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15603 2017-09-15  Richard Biener  <rguenther@suse.de>
15605         PR tree-optimization/82217
15606         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15607         but not undefined case.
15609 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15611         PR target/82145
15612         * postreload.c (reload_cse_simplify_operands): Skip
15613         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15615 2017-09-15  Richard Biener  <rguenther@suse.de>
15617         PR tree-optimization/68823
15618         * graphite-scop-detection.c (build_alias_set): If we have a
15619         possible dependence check whether we can handle them by just
15620         looking at the DRs DR_ACCESS_FNs.
15621         (build_scops): If build_alias_set fails, fail the SCOP.
15623 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15625         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15626         to support float128 built-in functions that require the ISA 3.0
15627         hardware.
15628         (BU_FLOAT128_3_HW): Likewise.
15629         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15630         built-in functions.
15631         (FMAF128): Likewise.
15632         (FMAQ): Likewise.
15633         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15634         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15635         floating point instructions.
15636         (rs6000_invalid_builtin): Likewise.
15637         (rs6000_builtin_mask_names): Likewise.
15638         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15639         (RS6000_BTM_FLOAT128_HW): Likewise.
15640         (RS6000_BTM_COMMON): Likewise.
15641         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15642         function.
15643         * doc/extend.texi (RS/6000 built-in functions): Document the
15644         IEEE 128-bit floating point square root and fused multiply-add
15645         built-in functions.
15647 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15649         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15650         reg (r2) isn't in the set of registers defined in the prologue.
15652 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15653             Alan Hayward  <alan.hayward@arm.com>
15654             David Sherwood  <david.sherwood@arm.com>
15656         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15657         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15658         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15659         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15660         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15661         accordingly.
15662         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15663         max_vectorization_factor.
15664         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15666 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15667             Alan Hayward  <alan.hayward@arm.com>
15668             David Sherwood  <david.sherwood@arm.com>
15670         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15671         (vect_worthwhile_without_simd_p): Declare.
15672         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15673         (vectorizable_reduction): Use it.
15674         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15675         (vectorizable_operation): Likewise.
15677 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15678             Alan Hayward  <alan.hayward@arm.com>
15679             David Sherwood  <david.sherwood@arm.com>
15681         * tree-vectorizer.h (vect_get_num_copies): New function.
15682         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15683         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15684         (vectorizable_induction): Likewise.
15685         (vectorizable_live_operation): Likewise.
15686         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15687         (vectorizable_bswap): Likewise.
15688         (vectorizable_call): Likewise.
15689         (vectorizable_conversion): Likewise.
15690         (vectorizable_assignment): Likewise.
15691         (vectorizable_shift): Likewise.
15692         (vectorizable_operation): Likewise.
15693         (vectorizable_store): Likewise.
15694         (vectorizable_load): Likewise.
15695         (vectorizable_condition): Likewise.
15696         (vectorizable_comparison): Likewise.
15697         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15699 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15700             Alan Hayward  <alan.hayward@arm.com>
15701             David Sherwood  <david.sherwood@arm.com>
15703         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15704         of vect_init_vector.
15706 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15707             Alan Hayward  <alan.hayward@arm.com>
15708             David Sherwood  <david.sherwood@arm.com>
15710         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15711         an inline wrapper that provides a location.
15712         (gimple_build_vector): Likewise.
15713         * gimple-fold.c (gimple_build_vector_from_val): New function.
15714         (gimple_build_vector): Likewise.
15715         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15716         functions to build the initial value.  Always return a gimple value.
15717         (get_initial_defs_for_reduction): Likewise.  Only compute
15718         neutral_vec once.
15719         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15720         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15721         (vectorizable_induction): Use gimple_build_vector rather than
15722         vect_init_vector.
15724 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15725             Alan Hayward  <alan.hayward@arm.com>
15726             David Sherwood  <david.sherwood@arm.com>
15728         * target.h (vec_perm_indices): New typedef.
15729         (auto_vec_perm_indices): Likewise.
15730         * optabs-query.h: Include target.h
15731         (can_vec_perm_p): Take a vec_perm_indices *.
15732         * optabs-query.c (can_vec_perm_p): Likewise.
15733         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15734         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15735         * tree-vect-generic.c (lower_vec_perm): Likewise.
15736         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15737         (vect_grouped_load_supported): Likewise.
15738         (vect_shift_permute_load_chain): Likewise.
15739         (vect_permute_store_chain): Use auto_vec_perm_indices.
15740         (vect_permute_load_chain): Likewise.
15741         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15742         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15743         Update uses of can_vec_perm_p.
15744         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15745         mode with a number of elements.  Take a vec_perm_indices *.
15746         (vect_create_epilog_for_reduction): Update accordingly.
15747         Use auto_vec_perm_indices.
15748         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15749         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15750         (vect_transform_slp_perm_load): Likewise.
15751         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15752         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15753         (vect_gen_perm_mask_checked): Likewise.
15754         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15755         (vect_gen_perm_mask_checked): Likewise.
15756         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15757         (vectorizable_store): Likewise.
15758         (vectorizable_load): Likewise.
15759         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15760         (vectorizable_bswap): Likewise.
15762 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15763             Alan Hayward  <alan.hayward@arm.com>
15764             David Sherwood  <david.sherwood@arm.com>
15766         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15767         * tree.c (build_vector): Likewise.
15768         (build_vector_from_ctor): Update accordingly.
15769         (build_vector_from_val): Likewise.
15770         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15771         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15772         * tree-vect-generic.c (add_rshift): Likewise.
15773         (expand_vector_divmod): Likewise.
15774         (optimize_vector_constructor): Likewise.
15775         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15776         (vect_transform_slp_perm_load): Likewise.
15777         (vect_schedule_slp_instance): Likewise.
15778         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15779         (vectorizable_call): Likewise.
15780         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15781         * expmed.c (make_tree): Likewise.
15782         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15783         a vector passed to build_vector.
15784         (fold_convert_const): Likewise.
15785         (exact_inverse): Likewise.
15786         (fold_ternary_loc): Likewise.
15787         (fold_relational_const): Likewise.
15788         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15789         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15790         (const_unop): Likewise.  Store the reduction accumulator in a
15791         variable rather than an array.
15792         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15793         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15794         the new vector, rather than constructing it after the input arrays.
15795         (native_interpret_vector): Use auto_vec<tree> when building
15796         a vector passed to build_vector.  Add elements in order.
15797         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15798         auto_vec<tree> when building a vector passed to build_vector.
15799         (vect_create_epilog_for_reduction): Likewise.
15800         (vectorizable_induction): Likewise.
15801         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15802         case statements.
15803         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15804         to a vec<tree> *.
15805         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15806         passed to build_vector.
15808 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15809             Alan Hayward  <alan.hayward@arm.com>
15810             David Sherwood  <david.sherwood@arm.com>
15812         * tree-core.h (tree_base::u): Add an "nelts" field.
15813         (tree_vector): Use VECTOR_CST_NELTS as the length.
15814         * tree.c (tree_size): Likewise.
15815         (make_vector): Initialize VECTOR_CST_NELTS.
15816         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15817         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15818         TYPE_VECTOR_SUBPARTS.
15819         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15820         as the number of units, setting it from VECTOR_CST_NELTS.
15821         (const_vector_from_tree): Likewise.
15822         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15823         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15824         (fold_negate_expr_1): Likewise.
15825         (fold_convert_const): Likewise.
15826         (const_binop): Likewise.  Differentiate the number of output and
15827         input elements.
15828         (const_unop): Likewise.
15829         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15830         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15831         in cases that did the opposite.
15833 2017-09-14  Richard Biener  <rguenther@suse.de>
15835         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15836         to VN_TOP.
15838 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15840         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15842 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15844         PR target/81325
15845         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15846         if and where to split a bb, except for splitting before debug insn
15847         sequences followed by non-label real insn.  Delete debug insns
15848         in between basic blocks.
15850         * combine.c (make_compound_operation_int): Formatting fixes.
15852         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15853         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15854         * config/netbsd.h (LINK_EH_SPEC): Likewise.
15855         * config/sol2.h (LINK_EH_SPEC): Likewise.
15856         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15857         * config/s390/linux.h (LINK_SPEC): Likewise.
15858         * config/freebsd.h (LINK_EH_SPEC): Likewise.
15859         * config/openbsd.h (LINK_EH_SPEC): Likewise.
15860         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15861         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15862         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
15863         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15864         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
15865         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
15867 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
15869         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
15870         support.
15871         (ENDFILE_LINUX_SPEC): Likewise.
15872         (LINK_EH_SPEC): Likewise.
15873         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
15874         (LINK_OS_LINUX_SPEC32): Likewise.
15875         (LINK_OS_LINUX_SPEC64): Likewise.
15876         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
15877         (LINK_OS_LINUX_SPEC): Likewise.
15879 2017-09-13  Martin Liska  <mliska@suse.cz>
15881         PR middle-end/82154
15882         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
15883         CASE_HIGH is NULL_TREE.
15885 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15886             Alan Hayward  <alan.hayward@arm.com>
15887             David Sherwood  <david.sherwood@arm.com>
15889         * target.def (secondary_memory_needed): New hook.
15890         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
15891         instead of SECONDARY_MEMORY_NEEDED.
15892         (secondary_memory_needed_mode): Likewise.
15893         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
15894         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
15895         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
15896         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
15897         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
15898         * doc/tm.texi: Regenerate.
15899         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
15900         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
15901         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15902         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
15903         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
15904         * config/i386/i386.c (inline_secondary_memory_needed): Put the
15905         mode argument first and change the reg_class arguments to reg_class_t.
15906         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
15907         Make static.  Update the call to inline_secondary_memory_needed.
15908         (ix86_register_move_cost): Update the call to
15909         inline_secondary_memory_needed.
15910         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15911         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
15912         definition.
15913         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
15914         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15915         in comment.
15916         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
15917         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
15918         * config/mips/mips.c (mips_secondary_memory_needed): Make static
15919         and match hook interface.  Add comment from mips.h.
15920         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15921         * config/mmix/mmix.md (truncdfsf2): Refer to
15922         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15923         in comment.
15924         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
15925         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
15926         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15927         (pa_secondary_memory_needed): New function.
15928         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
15929         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
15930         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15931         (pdp11_secondary_memory_needed): Make static and match hook interface.
15932         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
15933         * config/powerpcspe/powerpcspe-protos.h
15934         (rs6000_secondary_memory_needed_ptr): Delete.
15935         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
15936         Delete.
15937         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15938         (rs6000_option_override_internal): Assign to
15939         targetm.secondary_memory_needed rather than
15940         rs6000_secondary_memory_needed_ptr.
15941         (rs6000_secondary_memory_needed): Match hook interface.
15942         (rs6000_debug_secondary_memory_needed): Likewise.
15943         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
15944         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
15945         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
15946         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15947         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
15948         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
15949         Delete.
15950         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
15951         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15952         (rs6000_option_override_internal): Assign to
15953         targetm.secondary_memory_needed rather than
15954         rs6000_secondary_memory_needed_ptr.
15955         (rs6000_secondary_memory_needed): Match hook interface.
15956         (rs6000_debug_secondary_memory_needed): Likewise.
15957         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
15958         * config/s390/s390.c (s390_secondary_memory_needed): New function.
15959         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15960         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
15961         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15962         (sparc_secondary_memory_needed): New function.
15963         * lra-constraints.c (check_and_process_move): Refer to
15964         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15965         in comment.
15966         (curr_insn_transform): Likewise.
15967         (process_alt_operands): Use targetm.secondary_memory_needed
15968         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15969         (check_secondary_memory_needed_p): Likewise.
15970         (choose_split_class): Likewise.
15971         * reload.c: Unconditionally include code that was previously
15972         conditional on SECONDARY_MEMORY_NEEDED.
15973         (push_secondary_reload): Use targetm.secondary_memory_needed
15974         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15975         (push_reload): Likewise.
15976         * reload1.c: Unconditionally include code that was previously
15977         conditional on SECONDARY_MEMORY_NEEDED.
15978         (choose_reload_regs): Use targetm.secondary_memory_needed
15979         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15980         (gen_reload): Likewise.
15981         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
15983 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15984             Alan Hayward  <alan.hayward@arm.com>
15985             David Sherwood  <david.sherwood@arm.com>
15987         * target.def (secondary_memory_needed_mode): New hook:
15988         * targhooks.c (default_secondary_memory_needed_mode): Declare.
15989         * targhooks.h (default_secondary_memory_needed_mode): New function.
15990         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
15991         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
15992         * doc/tm.texi: Regenerate.
15993         * lra-constraints.c (check_and_process_move): Use
15994         targetm.secondary_memory_needed_mode instead of
15995         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
15996         (curr_insn_transform): Likewise.
15997         * reload.c (get_secondary_mem): Likewise.
15998         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15999         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16000         function.
16001         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16002         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16003         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16004         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16005         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16006         Delete.
16007         * config/powerpcspe/powerpcspe-protos.h
16008         (rs6000_secondary_memory_needed_mode): Delete.
16009         * config/powerpcspe/powerpcspe.c
16010         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16011         (rs6000_secondary_memory_needed_mode): Make static.
16012         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16013         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16014         Delete.
16015         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16016         Redefine.
16017         (rs6000_secondary_memory_needed_mode): Make static.
16018         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16019         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16020         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16021         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16022         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16023         Redefine.
16024         (sparc_secondary_memory_needed_mode): New function.
16025         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16027 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16029         * config/aarch64/constraints.md (Umq): New constraint.
16030         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16031         Change to use Umq.
16032         (mov<mode>): Update condition.
16034 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16036         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16037         when bitposition is the same.
16039 2017-09-13  Richard Biener  <rguenther@suse.de>
16041         * dwarf2out.c (output_die_symbol): Remove.
16042         (output_die): Do not output a DIEs symbol.
16044 2017-09-13  Richard Biener  <rguenther@suse.de>
16046         PR middle-end/82128
16047         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16048         default-def to avoid breaking iterator update with the weird
16049         interaction with cgraph_update_edges_for_call_stmt_node.
16051 2017-09-13  Richard Biener  <rguenther@suse.de>
16053         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16054         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16055         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16056         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16058 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16060         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16061         Disable pc relative literal load irrespective of
16062         TARGET_FIX_ERR_A53_84341 for default.
16064 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16066         * config/sparc/sparc.c (output_return): Output the source location of
16067         the insn in the delay slot, if any.
16068         (output_sibcall): Likewise.
16070 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16072         PR driver/81498
16073         * common.opt (-static-pie): New alias.
16074         (shared): Negate static-pie.
16075         (-no-pie): Update help text.
16076         (-pie): Likewise.
16077         (static-pie): New option.
16078         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16079         -static-pie support.
16080         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16081         (LINK_EH_SPEC): Likewise.
16082         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16083         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16084         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16085         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16086         (init_gcc_specs): Likewise.
16087         (init_spec): Likewise.
16088         (display_help): Update help message for -pie.
16089         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16090         -static-pie.
16092 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16094         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16095         (movdi_aarch64): Likewise.
16096         (movti_aarch64): Likewise.
16098 2017-09-12 Simon Wright <simon@pushface.org>
16100         PR target/80204
16101         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16102         calculation of the minor version, always output as 0.
16104 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16106         PR target/82112
16107         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16108         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16109         on it early, rather than manual conversion late.  For
16110         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16111         instead of performing manual conversion.
16113 2017-09-12  Carl Love  <cel@us.ibm.com>
16115         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16116         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16117         vmulouw, vmulosw.
16118         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16119         VMULOSW): Add definitions.
16120         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16121         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16122         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16123         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16124         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16126 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16128         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16129         types correctly.
16130         (movti_aarch64): Likewise.
16131         (movdf_aarch64): Likewise.
16132         (movtf_aarch64): Likewise.
16133         (load_pairdi): Likewise.
16134         (store_pairdi): Likewise.
16135         (load_pairdf): Likewise.
16136         (store_pairdf): Likewise.
16137         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16138         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16139         (ldr_got_small_<mode>): Likewise.
16140         (ldr_got_small_28k_<mode>): Likewise.
16141         (ldr_got_tiny): Likewise.
16142         * config/aarch64/iterators.md (ldst_sz): New.
16143         (ldpstp_sz): Likewise.
16144         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16145         to store_16.
16146         (thunderx_load): Split load_8 to load_16.
16147         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16148         load_8 to load_16.
16149         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16150         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16151         (xgene1_store_pair): Split store_8 to store_16.
16152         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16153         (falkor_st_0_st_sd): Split store_8 to store_16.
16155 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16157         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16158         and store1/2/3/4 to store_4/8/12/16.
16159         * config/aarch64/aarch64.md: Update for rename.
16160         * config/arm/arm.md: Likewise.: Likewise.
16161         * config/arm/arm.c: Likewise.
16162         * config/arm/thumb1.md: Likewise.
16163         * config/arm/thumb2.md: Likewise.
16164         * config/arm/vfp.md: Likewise.
16165         * config/arm/arm-generic.md: Likewise.
16166         * config/arm/arm1020e.md: Likewise.
16167         * config/arm/arm1026ejs.md: Likewise.
16168         * config/arm/arm1136jfs.md: Likewise.
16169         * config/arm/arm926ejs.md: Likewise.
16170         * config/arm/cortex-a15.md: Likewise.
16171         * config/arm/cortex-a17.md: Likewise.
16172         * config/arm/cortex-a5.md: Likewise.
16173         * config/arm/cortex-a53.md: Likewise.
16174         * config/arm/cortex-a57.md: Likewise.
16175         * config/arm/cortex-a7.md: Likewise.
16176         * config/arm/cortex-a8.md: Likewise.
16177         * config/arm/cortex-a9.md: Likewise.
16178         * config/arm/cortex-m4.md: Likewise.
16179         * config/arm/cortex-m7.md: Likewise.
16180         * config/arm/cortex-r4.md: Likewise.
16181         * config/arm/exynos-m1.md: Likewise.
16182         * config/arm/fa526.md: Likewise.
16183         * config/arm/fa606te.md: Likewise.
16184         * config/arm/fa626te.md: Likewise.
16185         * config/arm/fa726te.md: Likewise.
16186         * config/arm/fmp626.md: Likewise.
16187         * config/arm/iwmmxt.md: Likewise.
16188         * config/arm/ldmstm.md: Likewise.
16189         * config/arm/marvell-pj4.md: Likewise.
16190         * config/arm/xgene1.md: Likewise.
16191         * config/aarch64/thunderx.md: Likewise.
16192         * config/aarch64/thunderx2t99.md: Likewise.
16193         * config/aarch64/falkor.md: Likewise.
16195 2017-09-12  Martin Liska  <mliska@suse.cz>
16197         * attribs.c (private_lookup_attribute): New function.
16198         * attribs.h (private_lookup_attribute): Declared here.
16199         (lookup_attribute): Called from this place.
16201 2017-09-12  Richard Biener  <rguenther@suse.de>
16203         PR tree-optimization/82157
16204         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16205         stmts with side-effects.
16207 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16208             Alan Hayward  <alan.hayward@arm.com>
16209             David Sherwood <david.sherwood@arm.com>
16211         * target.def (hard_regno_nregs): New hook.
16212         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16213         * targhooks.h (default_hard_regno_nregs): Declare.
16214         * targhooks.c (default_hard_regno_nregs): New function.
16215         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16216         (TARGET_HARD_REGNO_NREGS): ...this hook.
16217         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16218         (CLASS_MAX_NREGS): Likewise.
16219         * doc/tm.texi: Regenerate.
16220         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16221         instead of HARD_REGNO_NREGS.
16222         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16223         HARD_REGNO_NREGS in the comment.
16224         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16225         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16226         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16227         Return an unsigned int.
16228         (TARGET_HARD_REGNO_NREGS): Redefine.
16229         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16230         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16231         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16232         (arc_hard_regno_nregs): New function.
16233         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16234         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16235         (arm_hard_regno_nregs): New function.
16236         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16237         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16238         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16239         (TARGET_HARD_REGNO_NREGS): Redefine.
16240         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16241         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16242         (HARD_REGNO_NREGS): Delete.
16243         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16244         (cr16_hard_regno_nregs): New function.
16245         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16246         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16247         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16248         (cris_hard_regno_nregs): New function.
16249         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16250         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16251         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16252         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16253         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16254         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16255         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16256         (frv_hard_regno_nregs): Make static.  Take and return an
16257         unsigned int.
16258         (frv_class_max_nregs): Remove outdated copy of documentation.
16259         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16260         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16261         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16262         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16263         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16264         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16265         (TARGET_HARD_REGNO_NREGS): Redefine.
16266         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16267         (CLASS_MAX_NREGS): Update comment.
16268         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16269         (ia64_hard_regno_nregs): New function.
16270         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16271         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16272         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16273         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16274         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16275         an unsigned int.
16276         (m32c_hard_regno_nregs): Likewise.  Make static.
16277         (TARGET_HARD_REGNO_NREGS): Redefine.
16278         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16279         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16280         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16281         (m68k_hard_regno_nregs): New function.
16282         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16283         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16284         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16285         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16286         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16287         Take and return an unsigned int.
16288         (TARGET_HARD_REGNO_NREGS): Redefine.
16289         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16290         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16291         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16292         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16293         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16294         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16295         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16296         (msp430_hard_regno_nregs): Make static.  Take and return an
16297         unsigned int.
16298         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16299         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16300         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16301         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16302         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16303         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16304         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16305         (TARGET_HARD_REGNO_NREGS): Redefine.
16306         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16307         (PA_HARD_REGNO_NREGS): ...this.
16308         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16309         (PA_HARD_REGNO_NREGS): ...this.
16310         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16311         (pa_hard_regno_nregs): New function.
16312         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16313         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16314         (pdp11_hard_regno_nregs): New function.
16315         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16316         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16317         (rs6000_hard_regno_nregs_hook): New function.
16318         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16319         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16320         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16321         Take and return an unsigned int.  Move earlier in file.
16322         (TARGET_HARD_REGNO_NREGS): Redefine.
16323         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16324         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16325         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16326         (rl78_hard_regno_nregs): Make static.  Take and return an
16327         unsigned int.
16328         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16329         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16330         (rs6000_hard_regno_nregs_hook): New function.
16331         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16332         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16333         (TARGET_HARD_REGNO_NREGS): Redefine.
16334         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16335         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16336         instead of HARD_REGNO_NREGS.
16337         (s390_hard_regno_nregs): New function.
16338         (s390_hard_regno_mode_ok): Add comment from s390.h.
16339         (TARGET_HARD_REGNO_NREGS): Redefine.
16340         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16341         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16342         (sh_hard_regno_nregs): New function.
16343         (sh_pass_in_reg_p): Use it.
16344         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16345         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16346         (sparc_hard_regno_nregs): New function.
16347         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16348         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16349         (spu_function_arg_advance): Use it, supplying a valid register number.
16350         (TARGET_HARD_REGNO_NREGS): Redefine.
16351         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16352         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16353         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16354         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16355         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16356         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16357         (CLASS_MAX_NREGS): Remove copy of old documentation.
16358         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16359         (visium_hard_regno_nregs): New function.
16360         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16361         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16362         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16363         xtensa_hard_regno_nregs): New function.
16364         * system.h (HARD_REGNO_NREGS): Poison.
16366 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16368         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16369         hard_regno_nregs instead of HARD_REGNO_NREGS.
16370         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16371         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16372         (c6x_expand_epilogue): Likewise.
16373         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16374         (frv_read_iacc_argument): Likewise.
16375         * config/sh/sh.c: Include regs.h.
16376         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16377         (regs_used): Likewise.
16378         (output_stack_adjust): Likewise.
16379         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16380         * expmed.c: Include regs.h.
16381         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16382         * ree.c: Include regs.h.
16383         (combine_reaching_defs): Use hard_regno_nregs instead of
16384         HARD_REGNO_NREGS.
16385         (add_removable_extension): Likewise.
16387 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16389         * regs.h (hard_regno_nregs): Turn into a function.
16390         (end_hard_regno): Update accordingly.
16391         * caller-save.c (setup_save_areas): Likewise.
16392         (save_call_clobbered_regs): Likewise.
16393         (replace_reg_with_saved_mem): Likewise.
16394         (insert_restore): Likewise.
16395         (insert_save): Likewise.
16396         * combine.c (can_change_dest_mode): Likewise.
16397         (move_deaths): Likewise.
16398         (distribute_notes): Likewise.
16399         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16400         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16401         (rs6000_split_multireg_move): Likewise.
16402         (rs6000_register_move_cost): Likewise.
16403         (rs6000_memory_move_cost): Likewise.
16404         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16405         (rs6000_split_multireg_move): Likewise.
16406         (rs6000_register_move_cost): Likewise.
16407         (rs6000_memory_move_cost): Likewise.
16408         * cselib.c (cselib_reset_table): Likewise.
16409         (cselib_lookup_1): Likewise.
16410         * emit-rtl.c (set_mode_and_regno): Likewise.
16411         * function.c (aggregate_value_p): Likewise.
16412         * ira-color.c (setup_profitable_hard_regs): Likewise.
16413         (check_hard_reg_p): Likewise.
16414         (calculate_saved_nregs): Likewise.
16415         (assign_hard_reg): Likewise.
16416         (improve_allocation): Likewise.
16417         (calculate_spill_cost): Likewise.
16418         * ira-emit.c (modify_move_list): Likewise.
16419         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16420         (ira_hard_reg_in_set_p): Likewise.
16421         * ira.c (setup_reg_mode_hard_regset): Likewise.
16422         (clarify_prohibited_class_mode_regs): Likewise.
16423         (check_allocation): Likewise.
16424         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16425         (lra_setup_reg_renumber): Likewise.
16426         (setup_try_hard_regno_pseudos): Likewise.
16427         (spill_for): Likewise.
16428         (assign_hard_regno): Likewise.
16429         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16430         * lra-constraints.c (in_class_p): Likewise.
16431         (lra_constraint_offset): Likewise.
16432         (simplify_operand_subreg): Likewise.
16433         (lra_constraints): Likewise.
16434         (split_reg): Likewise.
16435         (split_if_necessary): Likewise.
16436         (invariant_p): Likewise.
16437         (inherit_in_ebb): Likewise.
16438         * lra-lives.c (process_bb_lives): Likewise.
16439         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16440         (get_hard_regs): Likewise.
16441         (do_remat): Likewise.
16442         * lra-spills.c (assign_spill_hard_regs): Likewise.
16443         * mode-switching.c (create_pre_exit): Likewise.
16444         * postreload.c (reload_combine_recognize_pattern): Likewise.
16445         * recog.c (peep2_find_free_register): Likewise.
16446         * regcprop.c (kill_value_regno): Likewise.
16447         (set_value_regno): Likewise.
16448         (copy_value): Likewise.
16449         (maybe_mode_change): Likewise.
16450         (find_oldest_value_reg): Likewise.
16451         (copyprop_hardreg_forward_1): Likewise.
16452         * regrename.c (check_new_reg_p): Likewise.
16453         (regrename_do_replace): Likewise.
16454         * reload.c (push_reload): Likewise.
16455         (combine_reloads): Likewise.
16456         (find_dummy_reload): Likewise.
16457         (operands_match_p): Likewise.
16458         (find_reloads): Likewise.
16459         (find_equiv_reg): Likewise.
16460         (reload_adjust_reg_for_mode): Likewise.
16461         * reload1.c (count_pseudo): Likewise.
16462         (count_spilled_pseudo): Likewise.
16463         (find_reg): Likewise.
16464         (clear_reload_reg_in_use): Likewise.
16465         (free_for_value_p): Likewise.
16466         (allocate_reload_reg): Likewise.
16467         (choose_reload_regs): Likewise.
16468         (reload_adjust_reg_for_temp): Likewise.
16469         (emit_reload_insns): Likewise.
16470         (delete_output_reload): Likewise.
16471         * rtlanal.c (subreg_get_info): Likewise.
16472         * sched-deps.c (sched_analyze_reg): Likewise.
16473         * sel-sched.c (init_regs_for_mode): Likewise.
16474         (mark_unavailable_hard_regs): Likewise.
16475         (choose_best_reg_1): Likewise.
16476         (verify_target_availability): Likewise.
16477         * valtrack.c (dead_debug_insert_temp): Likewise.
16478         * var-tracking.c (track_loc_p): Likewise.
16479         (emit_note_insn_var_location): Likewise.
16480         * varasm.c (make_decl_rtl): Likewise.
16481         * reginfo.c (choose_hard_reg_mode): Likewise.
16482         (init_reg_modes_target): Refer directly to
16483         this_target_regs->x_hard_regno_nregs.
16485 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16487         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16488         instead of hard_regno_nregs.
16490 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16492         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16493         end_hard_regno instead of hard_regno_nregs.
16494         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16495         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16496         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16497         * ira-color.c (improve_allocation): Likewise.
16498         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16499         * lra-lives.c (mark_regno_live): Likewise.
16500         (mark_regno_dead): Likewise.
16501         * lra-remat.c (operand_to_remat): Likewise.
16502         * lra.c (collect_non_operand_hard_regs): Likewise.
16503         * postreload.c (reload_combine_note_store): Likewise.
16504         (move2add_valid_value_p): Likewise.
16505         * reload.c (regno_clobbered_p): Likewise.
16507 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16509         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16510         hard_regno_nregs.
16511         * config/v850/v850.c (v850_reorg): Likewise.
16512         * reload.c (refers_to_regno_for_reload_p): Likewise.
16513         (find_equiv_reg): Likewise.
16514         * reload1.c (reload_reg_reaches_end_p): Likewise.
16516 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16518         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16519         hard_regno_nregs.
16520         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16521         * config/arm/arm.c (output_move_neon): Likewise.
16522         (arm_attr_length_move_neon): Likewise.
16523         (neon_split_vcombine): Likewise.
16524         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16525         (c6x_mark_reg_written): Likewise.
16526         (c6x_dwarf_register_span): Likewise.
16527         * config/i386/i386.c (ix86_save_reg): Likewise.
16528         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16529         (rws_access_reg): Likewise.
16530         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16531         * mode-switching.c (create_pre_exit): Likewise.
16532         * ree.c (combine_reaching_defs): Likewise.
16533         (add_removable_extension): Likewise.
16534         * regcprop.c (find_oldest_value_reg): Likewise.
16535         (copyprop_hardreg_forward_1): Likewise.
16536         * reload.c (reload_inner_reg_of_subreg): Likewise.
16537         (push_reload): Likewise.
16538         (combine_reloads): Likewise.
16539         (find_dummy_reload): Likewise.
16540         (reload_adjust_reg_for_mode): Likewise.
16541         * reload1.c (find_reload_regs): Likewise.
16542         (forget_old_reloads_1): Likewise.
16543         (reload_reg_free_for_value_p): Likewise.
16544         (reload_adjust_reg_for_temp): Likewise.
16545         (emit_reload_insns): Likewise.
16546         (delete_output_reload): Likewise.
16547         * sel-sched.c (choose_best_reg_1): Likewise.
16548         (choose_best_pseudo_reg): Likewise.
16550 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16551             Alan Hayward  <alan.hayward@arm.com>
16552             David Sherwood <david.sherwood@arm.com>
16554         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16555         * target.def (slow_unaligned_access): New hook.
16556         * targhooks.h (default_slow_unaligned_access): Declare.
16557         * targhooks.c (default_slow_unaligned_access): New function.
16558         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16559         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16560         * doc/tm.texi: Regenerate.
16561         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16562         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16563         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16564         definition.
16565         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16566         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16567         Redefine.
16568         (rs6000_slow_unaligned_access): New function.
16569         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16570         (expand_block_compare): Likewise.
16571         (expand_strn_compare): Likewise.
16572         (rs6000_rtx_costs): Likewise.
16573         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16574         (riscv_slow_unaligned_access): Likewise.
16575         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16576         (riscv_slow_unaligned_access_p): ...this and make static.
16577         (riscv_option_override): Update accordingly.
16578         (riscv_slow_unaligned_access): New function.
16579         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16580         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16581         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16582         (rs6000_slow_unaligned_access): New function.
16583         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16584         (rs6000_rtx_costs): Likewise.
16585         * config/rs6000/rs6000-string.c (expand_block_compare)
16586         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16587         of SLOW_UNALIGNED_ACCESS.
16588         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16589         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16590         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16591         of SLOW_UNALIGNED_ACCESS.
16592         * expmed.c (simple_mem_bitfield_p): Likewise.
16593         * expr.c (alignment_for_piecewise_move): Likewise.
16594         (emit_group_load_1): Likewise.
16595         (emit_group_store): Likewise.
16596         (copy_blkmode_from_reg): Likewise.
16597         (emit_push_insn): Likewise.
16598         (expand_assignment): Likewise.
16599         (store_field): Likewise.
16600         (expand_expr_real_1): Likewise.
16601         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16602         * lra-constraints.c (simplify_operand_subreg): Likewise.
16603         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16604         * gimple-ssa-store-merging.c: Likewise in block comment at start
16605         of file.
16606         * tree-ssa-strlen.c: Include target.h.
16607         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16608         of SLOW_UNALIGNED_ACCESS.
16609         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16611 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16613         PR rtl-optimization/82185
16614         * expmed.c (emit_store_flag_int): Only test tem if it has been
16615         initialized.
16617 2017-09-12  Richard Biener  <rguenther@suse.de>
16619         PR middle-end/82149
16620         * match.pd ((FTYPE) N CMP CST): Fix typo.
16622 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16624         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16625         attribute.
16626         (mips_near_type_p): Add 'short_call' attribute as a synonym
16627         for 'near'.
16628         * doc/extend.texi (short_call): Document new function attribute.
16630 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16632         PR target/82112
16633         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16634         assertion check that in the condition.
16635         (get_atomic_generic_size): Likewise.  Before testing if parameter
16636         has pointer type, if it has array type, call for C++
16637         default_conversion to perform array-to-pointer conversion.
16639 2017-09-12  Richard Biener  <rguenther@suse.de>
16641         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16642         for operations we cannot scalarize.
16644 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16646         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16647         vectors heap vectors.  Clean up comments.
16648         Make visited_bbs a reference.
16649         (profitable_jump_thread_path): Make GC
16650         vectors heap vectors.  Clean up comments.
16651         Misc cleanups.
16652         (convert_and_register_jump_thread_path): Make GC vectors heap
16653         vectors.
16654         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16655         Make visited_bbs a reference.
16656         (handle_phi): Abstract common code to to
16657         register_jump_thread_path_if_profitable.
16658         Rename VAR_BB to DEF_BB.
16659         Update comments.
16660         Make GC vectors heap vectors.
16661         Make visited_bbs a reference.
16662         (handle_assignment): Same.
16663         (register_jump_thread_path_if_profitable): New.
16664         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16665         DEF_BB.
16666         Make GC vectors heap vectors.  Clean up comments.
16667         Make visited_bbs a reference.
16668         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16669         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16670         comment.
16672 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16674         PR target/82181
16675         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16676         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16678 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16680         Revert r251800 and r251799.
16682 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16684         PR hsa/82119
16685         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16686         arguments in advance.
16687         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16688         use it to find predecessor edges.
16689         (naive_outof_ssa): Collect vector of predecessors.
16691 2017-09-08  Jason Merrill  <jason@redhat.com>
16693         PR c++/70029 - ICE with ref-qualifier and -flto
16694         * langhooks.h (struct lang_hooks_for_types): Add
16695         copy_lang_qualifiers.
16696         * attribs.c (build_type_attribute_qual_variant): Use it.
16697         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16698         NULL.
16699         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16700         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16702 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16704         PR target/81988
16705         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16706         (*mulsi3_sp64): New instruction.
16707         (mulsi3): New expander.
16709 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16711         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16713 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16715         * sancov.c: Include memmodel.h.
16717 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16719         PR target/80897
16720         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16721         large offsets.
16723 2017-09-07  Carl Love  <cel@us.ibm.com>
16725         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16726         the sldi instruction.
16728 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16730         * sancov.c: Include tm_p.h.
16732 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16734         PR target/81979
16735         * output.h (switch_to_other_text_partition): New declaration.
16736         * varasm.c (switch_to_other_text_partition): New function.
16737         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16738         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16739         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16740         to the other text partition before emitting LCL label and switch back
16741         after emitting the word after it.
16743 2017-09-07  Richard Biener  <rguenther@suse.de>
16745         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16746         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16747         critical edges here, after loop init.
16748         (pass_data_pre): Remove PROP_no_crit_edges flags.
16749         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16750         for valueization of call args to avoid leaking VN_TOP.
16751         (visit_use): Assert we do not visit default defs.
16752         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16753         Use error_mark_node to more easily detect leaking VN_TOP.
16754         All default-defs are varying, not VN_TOP.  Mark them visited.
16755         (run_scc_vn): Make code match comment.
16757 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16759         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16760         OPTION_MASK_FLOAT128_KEYWORD.
16761         (POWERPC_MASKS): Likewise.
16762         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16763         support for the -mfloat128-type option, and make -mfloat128
16764         default on PowerPC Linux systems.  Define or undefine
16765         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16766         Define __float128 to be __ieee128 if IEEE 128-bit support is
16767         enabled, or undefine it.
16768         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16769         Delete defining __FLOAT128_TYPE__.
16770         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16771         -mfloat128-type option and make -mfloat128 default on PowerPC
16772         Linux systems.
16773         (TARGET_FLOAT128_TYPE): Likewise.
16774         (-mfloat128-type): Likewise.
16775         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16776         Delete the -mfloat128-type option and make -mfloat128 default on
16777         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16778         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16779         128-bit floating point is enabled.  Change tests from using
16780         -mfloat128-type to -mfloat128.
16781         (rs6000_mangle_type): Use the correct mangling for the __float128
16782         type even if normal long double is restricted to 64-bits.
16783         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16784         systems.
16785         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16786         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16787         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16788         -mfloat128-type.
16789         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16790         documentation for -mfloat128.
16792 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16794         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16796 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16797             Jakub Jelinek  <jakub@redhat.com>
16799         * asan.c (initialize_sanitizer_builtins): Add
16800         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16801         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16802         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16803         BT_FN_VOID_UINT64_PTR variables.
16804         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16805         (BT_FN_VOID_UINT16_UINT16): Likewise.
16806         (BT_FN_VOID_UINT32_UINT32): Likewise.
16807         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16808         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16809         (BT_FN_VOID_UINT64_PTR): Likewise.
16810         * common.opt (flag_sanitize_coverage): New variable.
16811         (fsanitize-coverage=trace-pc): Remove.
16812         (fsanitize-coverage=): Add.
16813         * flag-types.h (enum sanitize_coverage_code): New enum.
16814         * fold-const.c (fold_range_test): Disable non-short-circuit
16815         optimization if flag_sanitize_coverage.
16816         (fold_truth_andor): Likewise.
16817         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16818         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16819         (coverage_sanitizer_opts): New array.
16820         (get_closest_sanitizer_option): Add OPTS argument, handle also
16821         OPT_fsanitize_coverage_.
16822         (parse_sanitizer_options): Adjusted to also handle
16823         OPT_fsanitize_coverage_.
16824         (common_handle_option): Add OPT_fsanitize_coverage_.
16825         * sancov.c (instrument_comparison, instrument_switch): New function.
16826         (sancov_pass): Add trace-cmp support.
16827         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16828         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16829         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16830         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16831         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16832         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16833         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16834         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16835         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16836         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16838 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16840         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16841         error.  Only quit immediately if parsing is complete.
16842         (BEGIN): Initialize fatal_err and parse_done.
16843         (begin fpu, end fpu): Check number of arguments.
16844         (begin arch, end arch): Likewise.
16845         (begin cpu, end cpu): Likewise.
16846         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16847         (optalias): Likewise.
16849 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16851         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16852         * config/arm/arm-isa.h: Delete.  Move definitions to ...
16853         * arm-cpus.in: ... here.  Use new feature and fgroup values.
16854         * config/arm/arm.c (arm_option_override): Use lower case for feature
16855         bit names.
16856         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16857         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16858         * config/arm/parsecpu.awk (END): Add new command 'isa'.
16859         (isa_pfx): Delete.
16860         (print_isa_bits_for): New function.
16861         (gen_isa): New function.
16862         (gen_comm_data): Use print_isa_bits_for.
16863         (define feature): New keyword.
16864         (define fgroup): New keyword.
16865         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
16866         (arm-isa.h): Add rule to generate file.
16867         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
16868         case for feature bit names.
16870 2017-09-06  Richard Biener  <rguenther@suse.de>
16872         * tree-ssa-pre.c (NECESSARY): Remove.
16873         (create_expression_by_pieces): Do not touch pass-local flags.
16874         (insert_into_preds_of_block): Likewise.
16875         (do_pre_regular_insertion): Likewise.
16876         (eliminate_insert): Likewise.
16877         (eliminate_dom_walker::before_dom_children): Likewise.
16878         (fini_eliminate): Do not look at inserted_exprs.
16879         (mark_operand_necessary): Remove.
16880         (remove_dead_inserted_code): Replace with simple work-list
16881         algorithm based on inserted_exprs and SSA uses.
16882         (pass_pre::execute): Re-order fini_eliminate and
16883         remove_dead_inserted_code.
16885 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16887         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
16888         for VxWorks 7.  Adjust surrounding comments.
16890 2017-09-06  Richard Biener  <rguenther@suse.de>
16892         * gimple-ssa-strength-reduction.c
16893         (find_candidates_dom_walker::before_dom_children): Also allow
16894         pointer types.
16896 2017-09-06  Richard Biener  <rguenther@suse.de>
16898         PR tree-optimization/82108
16899         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
16900         for gap in the non-permutation SLP case.
16902 2017-09-06  Martin Jambor  <mjambor@suse.cz>
16904         PR tree-optimization/82078
16905         * tree-sra.c (sort_and_splice_var_accesses): Move call to
16906         add_access_to_work_queue...
16907         (build_accesses_from_assign): ...here.
16908         (propagate_all_subaccesses): Make sure racc is the group
16909         representative, if there is one.
16911 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
16913         PR middle-end/82095
16914         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
16915         NULL DECL_INITIAL.
16917 2017-09-06  Richard Biener  <rguenther@suse.de>
16919         * gimple-ssa-strength-reduction.c
16920         (find_candidates_dom_walker::before_doom_children): Use a
16921         type and not a mode check.
16923 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16925         PR target/77308
16926         * config/arm/predicates.md (arm_general_adddi_operand): Create new
16927         non-vfp predicate.
16928         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
16930 2017-09-05  Jeff Law  <law@redhat.com>
16932         PR tree-optimization/64910
16933         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
16934         cases where we have 3 or more operands.
16936 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
16938         PR middle-end/81768
16939         * omp-low.c (lower_omp_for): Recompute tree invariant if
16940         gimple_omp_for_initial/final is ADDR_EXPR.
16942         PR middle-end/81768
16943         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
16944         into gimple val before gimplification fo the COND_EXPR.
16946 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
16948         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
16949         REGION_COPY argument.
16950         (thread_through_all_blocks): Remove unused argument to
16951         duplicate_thread_path.
16953 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16954             Alan Hayward  <alan.hayward@arm.com>
16955             David Sherwood  <david.sherwood@arm.com>
16957         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
16958         Take a scalar_mode rather than a machine_mode.
16959         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16960         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
16961         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16962         (aarch64_gen_adjusted_ldpstp): Likewise.
16963         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
16965 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16966             Alan Hayward  <alan.hayward@arm.com>
16967             David Sherwood  <david.sherwood@arm.com>
16969         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
16970         Take a scalar_int_mode instead of a machine_mode.
16971         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
16972         (aarch64_output_scalar_simd_mov_immediate): Likewise.
16973         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
16974         (aarch64_simd_attr_length_rglist): Delete.
16975         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
16976         a scalar_int_mode instead of a machine_mode.
16977         (aarch64_add_offset): Likewise.
16978         (aarch64_internal_mov_immediate): Likewise
16979         (aarch64_add_constant_internal): Likewise.
16980         (aarch64_add_constant): Likewise.
16981         (aarch64_movw_imm): Likewise.
16982         (aarch64_rtx_arith_op_extract_p): Likewise.
16983         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
16984         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
16985         Remove assert that the mode isn't a vector.
16986         (aarch64_output_scalar_simd_mov_immediate): Likewise.
16987         (aarch64_expand_mov_immediate): Update calls after above changes.
16988         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
16989         (aarch64_and_bitmask_imm): Check for scalar integer modes.
16990         (aarch64_move_imm): Likewise.
16991         (aarch64_can_const_movi_rtx_p): Likewise.
16992         (aarch64_strip_extend): Likewise.
16993         (aarch64_extr_rtx_p): Likewise.
16994         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
16995         a CONST_INT when the mode parameter is VOIDmode.
16996         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
16998 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17000         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17001         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17002         (bitwise_type_for_mode): Update accordingly.
17004 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17006         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17007         * stor-layout.c (mode_for_size_tree): Likewise.
17008         (mode_for_array): Update accordingly.
17009         (layout_decl): Likewise.
17010         (compute_record_mode): Likewise.  Only set the mode once.
17012 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17014         * target.def (get_mask_mode): Change return type to opt_mode.
17015         Expand commentary.
17016         * doc/tm.texi: Regenerate.
17017         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17018         * targhooks.c (default_get_mask_mode): Likewise.
17019         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17020         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17021         targetm.get_mask_mode.
17022         * tree.c (build_truth_vector_type): Likewise.
17024 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17026         * machmode.h (mode_for_vector): Return an opt_mode.
17027         * stor-layout.c (mode_for_vector): Likewise.
17028         (mode_for_int_vector): Update accordingly.
17029         (layout_type): Likewise.
17030         * config/i386/i386.c (emit_memmov): Likewise.
17031         (ix86_expand_set_or_movmem): Likewise.
17032         (ix86_expand_vector_init): Likewise.
17033         (ix86_get_mask_mode): Likewise.
17034         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17035         Likewise.
17036         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17037         * expmed.c (extract_bit_field_1): Likewise.
17038         * expr.c (expand_expr_real_2): Likewise.
17039         * optabs-query.c (can_vec_perm_p): Likewise.
17040         (can_vec_mask_load_store_p): Likewise.
17041         * optabs.c (expand_vec_perm): Likewise.
17042         * targhooks.c (default_get_mask_mode): Likewise.
17043         * tree-vect-stmts.c (vectorizable_store): Likewise.
17044         (vectorizable_load): Likewise.
17045         (get_vectype_for_scalar_type_and_size): Likewise.
17047 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17049         * machmode.h (mode_for_int_vector): New function.
17050         * stor-layout.c (mode_for_int_vector): Likewise.
17051         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17052         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17053         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17054         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17055         (s390_expand_vcond): Likewise.
17057 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17059         * machmode.h (opt_machine_mode): New type.
17060         (opt_mode<T>): Allow construction from anything that can be
17061         converted to a T.
17062         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17063         (mode_for_size): Return an opt_machine_mode.
17064         * stor-layout.c (mode_for_size): Likewise.
17065         (mode_for_size_tree): Update call accordingly.
17066         (bitwise_mode_for_mode): Likewise.
17067         (make_fract_type): Likewise.
17068         (make_accum_type): Likewise.
17069         * caller-save.c (replace_reg_with_saved_mem): Update call
17070         accordingly.
17071         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17072         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17073         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17074         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17075         * expmed.c (extract_bit_field_1): Likewise.
17076         * reload.c (get_secondary_mem): Likewise.
17077         * varasm.c (assemble_integer): Likewise.
17078         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17079         early-out.
17081 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17083         * machmode.h (decimal_float_mode_for_size): New function.
17084         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17085         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17086         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17087         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17088         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17089         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17091 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17093         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17094         (get_builtin_sync_mode): Likewise.
17095         (expand_ifn_atomic_compare_exchange): Likewise.
17096         (expand_builtin_atomic_clear): Likewise.
17097         (expand_builtin_atomic_test_and_set): Likewise.
17098         (fold_builtin_atomic_always_lock_free): Likewise.
17099         * calls.c (compute_argument_addresses): Likewise.
17100         (emit_library_call_value_1): Likewise.
17101         (store_one_arg): Likewise.
17102         * combine.c (combine_instructions): Likewise.
17103         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17104         * config/arm/arm.c (arm_function_value): Likewise.
17105         (aapcs_allocate_return_reg): Likewise.
17106         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17107         * config/i386/i386.c (construct_container): Likewise.
17108         (ix86_gimplify_va_arg): Likewise.
17109         (ix86_expand_sse_cmp): Likewise.
17110         (emit_memmov): Likewise.
17111         (emit_memset): Likewise.
17112         (expand_small_movmem_or_setmem): Likewise.
17113         (ix86_expand_pextr): Likewise.
17114         (ix86_expand_pinsr): Likewise.
17115         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17116         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17117         Likewise.
17118         * config/mips/mips.c (mips_function_value_1) Likewise.
17119         (mips_block_move_straight): Likewise.
17120         (mips_expand_ins_as_unaligned_store): Likewise.
17121         * config/powerpcspe/powerpcspe.c
17122         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17123         (rs6000_darwin64_record_arg_flush): Likewise.
17124         * config/rs6000/rs6000.c
17125         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17126         (rs6000_darwin64_record_arg_flush): Likewise.
17127         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17128         (sparc_function_value_1): Likewise.
17129         * config/spu/spu.c (adjust_operand): Likewise.
17130         (spu_emit_branch_or_set): Likewise.
17131         (arith_immediate_p): Likewise.
17132         * emit-rtl.c (gen_lowpart_common): Likewise.
17133         * expr.c (expand_expr_real_1): Likewise.
17134         * function.c (assign_parm_setup_block): Likewise.
17135         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17136         * reload1.c (alter_reg): Likewise.
17137         * stor-layout.c (mode_for_vector): Likewise.
17138         (layout_type): Likewise.
17140 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17142         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17143         (spu_convert_move): Likewise.
17144         * lower-subreg.c (resolve_simple_move): Likewise.
17146 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17148         PR target/81833
17149         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17150         define_insn to a define_expand.
17151         (altivec_vsum2sws_direct): New define_insn.
17152         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17154 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17156         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17157         max_insns_skipped.
17159 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17161         PR target/59501
17162         PR target/81624
17163         PR target/81769
17164         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17165         realign stack if stack alignment needed is less than incoming
17166         stack boundary.
17168 2017-09-05  Marek Polacek  <polacek@redhat.com>
17170         PR sanitizer/82072
17171         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17172         check earlier.
17174 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17176         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17178 2017-09-05  Richard Biener  <rguenther@suse.de>
17180         PR tree-optimization/82084
17181         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17183 2017-09-05  Richard Biener  <rguenther@suse.de>
17185         PR tree-optimization/82102
17186         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17188 2017-09-05  Martin Liska  <mliska@suse.cz>
17190         PR tree-optimization/82032
17191         * tree-cfg.c (generate_range_test): New function.
17192         * tree-cfg.h (generate_range_test): Declared here.
17193         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17194         (cleanup_control_expr_graph): Use it.
17195         * tree-switch-conversion.c (try_switch_expansion): Remove
17196         assert.
17197         (emit_case_nodes): Use generate_range_test.
17199 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17201         PR target/82098
17202         * config/i386/i386.md (*<btsc><mode>_mask): Add
17203         TARGET_USE_BT to insn constraint.
17204         (*btr<mode>_mask): Ditto.
17206 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17208         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17209         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17211 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17213         PR target/77308
17214         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17215         TARGET_NEON and TARGET_IWMMXT.
17216         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17217         TARGET_NEON and TARGET_IWMMXT.
17218         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17220 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17222         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17223         (ix86_rewrite_tls_address): Ditto.
17224         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17225         (ix86_rewrite_tls_address_1): Ditto.
17226         (ix86_rewrite_tls_address): Ditto.
17227         * config/i386/predicates.md (tls_address_pattern): New predicate.
17228         * config/i386/i386.md (TLS address splitter): New splitter.
17230 2017-09-04  Richard Biener  <rguenther@suse.de>
17232         PR tree-optimization/82084
17233         * fold-const.h (can_native_encode_string_p): Declare.
17234         * fold-const.c (can_native_encode_string_p): Factor out from ...
17235         (native_encode_string): ... here.
17236         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17237         vectorizing stores from constants we later cannot handle.
17239 2017-09-04  Marek Polacek  <polacek@redhat.com>
17241         PR c/81783
17242         * doc/invoke.texi: Update -Wtautological-compare documentation.
17244 2017-09-04  Jeff Law  <law@redhat.com>
17246         PR tree-optimization/64910
17247         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17248         swap the first and last operand if the last is a constant.
17250 2017-09-04  Marek Polacek  <polacek@redhat.com>
17252         PR sanitizer/82072
17253         * convert.c (do_narrow): When sanitizing signed integer overflows,
17254         bail out for signed types.
17255         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17257 2017-09-04  Richard Biener  <rguenther@suse.de>
17259         PR tree-optimization/82060
17260         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17261         Move devirtualization after stmt folding and before EH/AB/noreturn
17262         cleanup to get the stmt refs canonicalized.  Use a bool instead
17263         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17264         NOPs generated by folding for removal.
17266 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17267             Alan Hayward  <alan.hayward@arm.com>
17268             David Sherwood  <david.sherwood@arm.com>
17270         * coretypes.h (pad_direction): New enum.
17271         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17272         (FUNCTION_ARG_PADDING): Likewise.
17273         * target.def (function_arg_padding): New hook.
17274         * targhooks.h (default_function_arg_padding): Declare.
17275         * targhooks.c (default_function_arg_padding): New function.
17276         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17277         (TARGET_FUNCTION_ARG_PADDING): ...this.
17278         * doc/tm.texi: Regenerate.
17279         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17280         instead of direction.
17281         (compute_argument_addresses): Likewise.
17282         (load_register_parameters): Likewise.
17283         (emit_library_call_value_1): Likewise.
17284         (store_one_arg): Use targetm.calls.function_arg_padding instead
17285         of FUNCTION_ARG_PADDING.
17286         (must_pass_in_stack_var_size_or_pad): Likewise.
17287         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17288         (emit_group_store): Likewise.
17289         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17290         instead of FUNCTION_ARG_PADDING.
17291         (emit_push_insn): Likewise, and propagate enum change throughout
17292         function.
17293         * function.h (direction): Delete.
17294         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17295         of direction.
17296         * function.c (assign_parm_find_stack_rtl): Likewise.
17297         (assign_parm_setup_block_p): Likewise.
17298         (assign_parm_setup_block): Likewise.
17299         (gimplify_parameters): Likewise.
17300         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17301         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17302         function.
17303         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17304         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17305         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17306         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17307         (aarch64_function_arg_padding): ...this new function.
17308         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17309         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17310         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17311         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17312         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17313         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17314         (arm_pad_arg_upward): Replace with...
17315         (arm_function_arg_padding): ...this new function.
17316         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17317         of direction.
17318         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17319         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17320         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17321         (ia64_hpux_function_arg_padding): Replace with...
17322         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17323         instead of direction.  Check for TARGET_HPUX.
17324         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17325         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17326         (iq2000_function_arg_padding): New function.
17327         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17328         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17329         (mips_function_arg_padding): ...this new function.
17330         (mips_pad_reg_upward): Update accordingly.
17331         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17332         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17333         targetm.calls.function_arg_padding.
17334         (FUNCTION_ARG_PADDING): Delete.
17335         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17336         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17337         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17338         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17339         (nios2_block_reg_padding): Return pad_direction instead of direction.
17340         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17341         instead of direction.
17342         (nios2_function_arg_padding): Likewise.  Make static.
17343         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17344         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17345         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17346         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17347         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17348         (pa_function_arg_padding): Make static.  Return pad_direction instead
17349         of direction.
17350         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17351         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17352         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17353         instead of direction.  Use targetm.calls.function_arg_padding.
17354         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17355         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17356         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17357         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17358         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17359         Redefine.
17360         (function_arg_padding): Rename to...
17361         (rs6000_function_arg_padding): ...this.  Make static.  Return
17362         pad_direction instead of direction.
17363         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17364         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17365         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17366         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17367         instead of direction.  Use targetm.calls.function_arg_padding.
17368         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17369         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17370         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17371         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17372         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17373         (function_arg_padding): Rename to...
17374         (rs6000_function_arg_padding): ...this.  Make static.  Return
17375         pad_direction instead of direction.
17376         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17377         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17378         * config/s390/s390.c (s390_function_arg_padding): New function.
17379         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17380         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17381         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17382         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17383         (function_arg_padding): Rename to...
17384         (sparc_function_arg_padding): ...this.  Make static.  Return
17385         pad_direction instead of direction.
17386         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17387         * config/spu/spu.c (spu_function_arg_padding): New function.
17388         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17389         * system.h (FUNCTION_ARG_PADDING): Poison.
17391 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17392             Alan Hayward  <alan.hayward@arm.com>
17393             David Sherwood  <david.sherwood@arm.com>
17395         * target.def (modes_tieable_p): New hook.
17396         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17397         (TARGET_MODES_TIEABLE_P): ...this.
17398         * doc/tm.texi.in: Regenerate.
17399         * hooks.h (hook_bool_mode_mode_true): Declare.
17400         * hooks.c (hook_bool_mode_mode_true): New function.
17401         * combine.c (subst): Use targetm.modes_tieable_p instead of
17402         MODES_TIEABLE_P.
17403         * dse.c (find_shift_sequence): Likewise.
17404         * expmed.c (extract_low_bits): Likewise.
17405         * lower-subreg.c: Include target.h.
17406         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17407         MODES_TIEABLE_P.
17408         * rtlanal.c (rtx_cost): Likewise.
17409         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17410         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17411         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17412         (TARGET_MODES_TIEABLE_P): Redefine.
17413         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17414         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17415         (TARGET_MODES_TIEABLE_P): Redefine.
17416         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17417         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17418         (arc_modes_tieable_p): New function.
17419         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17420         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17421         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17422         (arm_modes_tieable_p): Make static.
17423         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17424         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17425         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17426         (TARGET_MODES_TIEABLE_P): Redefine.
17427         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17428         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17429         (TARGET_MODES_TIEABLE_P): Redefine.
17430         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17431         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17432         (cr16_modes_tieable_p): New function.
17433         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17434         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17435         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17436         (TRULY_NOOP_TRUNCATION): Update comment.
17437         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17438         (TRULY_NOOP_TRUNCATION): Update comment.
17439         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17440         (frv_modes_tieable_p): New function.
17441         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17442         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17443         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17444         (TARGET_MODES_TIEABLE_P): Redefine.
17445         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17446         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17447         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17448         (TARGET_MODES_TIEABLE_P): Redefine.
17449         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17450         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17451         (ia64_modes_tieable_p): New function.
17452         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17453         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17454         (iq2000_modes_tieable_p): New function.
17455         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17456         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17457         (lm32_modes_tieable_p): New function.
17458         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17459         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17460         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17461         (TARGET_MODES_TIEABLE_P): Redefine.
17462         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17463         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17464         (m32r_modes_tieable_p): New function.
17465         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17466         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17467         (m68k_modes_tieable_p): New function.
17468         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17469         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17470         (mcore_modes_tieable_p): New function.
17471         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17472         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17473         function.
17474         (TARGET_MODES_TIEABLE_P): Redefine.
17475         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17476         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17477         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17478         (TARGET_MODES_TIEABLE_P): Redefine.
17479         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17480         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17481         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17482         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17483         (mn10300_modes_tieable_p): ...this and make static.
17484         (TARGET_MODES_TIEABLE_P): Redefine.
17485         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17486         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17487         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17488         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17489         (msp430_modes_tieable_p): Make static.
17490         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17491         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17492         (TARGET_MODES_TIEABLE_P): Redefine.
17493         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17494         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17495         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17496         (TARGET_MODES_TIEABLE_P): Redefine.
17497         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17498         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17499         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17500         (TARGET_MODES_TIEABLE_P): Redefine.
17501         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17502         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17503         (pdp11_modes_tieable_p): New function.
17504         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17505         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17506         (rs6000_modes_tieable_p): New function.
17507         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17508         * config/powerpcspe/powerpcspe.md: Update comment.
17509         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17510         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17511         (TARGET_MODES_TIEABLE_P): Redefine.
17512         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17513         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17514         (rl78_modes_tieable_p): New function.
17515         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17516         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17517         (rs6000_modes_tieable_p): New function.
17518         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17519         * config/rs6000/rs6000.md: Update comment.
17520         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17521         * config/rx/rx.c (rx_modes_tieable_p): New function.
17522         (TARGET_MODES_TIEABLE_P): Redefine.
17523         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17524         * config/s390/s390.c (s390_modes_tieable_p): New function.
17525         (TARGET_MODES_TIEABLE_P): Redefine.
17526         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17527         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17528         (sh_modes_tieable_p): New function.
17529         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17530         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17531         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17532         (sparc_modes_tieable_p): Make static.
17533         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17534         * config/spu/spu.c (spu_modes_tieable_p): New function.
17535         (TARGET_MODES_TIEABLE_P): Redefine.
17536         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17537         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17538         (TARGET_MODES_TIEABLE_P): Redefine.
17539         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17540         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17541         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17542         * config/v850/v850.c (v850_modes_tieable_p): New function.
17543         (TARGET_MODES_TIEABLE_P): Redefine.
17544         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17545         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17546         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17547         (visium_modes_tieable_p): New function.
17548         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17549         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17550         (xtensa_modes_tieable_p): New function.
17551         * system.h (MODES_TIEABLE_P): Poison.
17553 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17554             Alan Hayward  <alan.hayward@arm.com>
17555             David Sherwood  <david.sherwood@arm.com>
17557         * target.def (hard_regno_mode_ok): New hook.
17558         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17559         (TARGET_HARD_REGNO_MODE_OK): ...this.
17560         * doc/tm.texi.in: Regenerate.
17561         * hooks.h (hook_bool_uint_mode_true): Declare.
17562         * hooks.c (hook_bool_uint_mode_true): New function.
17563         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17564         HARD_REGNO_MODE_OK.
17565         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17566         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17567         instead of HARD_REGNO_MODE_OK.
17568         * caller-save.c: Include target.h.
17569         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17570         HARD_REGNO_MODE_OK.
17571         * combine.c (can_combine_p): Likewise.
17572         (combinable_i3pat): Likewise.
17573         (can_change_dest_mode): Likewise.
17574         * expr.c (init_expr_target): Likewise.
17575         (convert_move): Likewise.
17576         (convert_modes): Likewise.
17577         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17578         (setup_prohibited_mode_move_regs): Likewise.
17579         * ira.h (target_ira): Likewise.
17580         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17581         * lra-constraints.c (process_alt_operands): Likewise.
17582         (split_reg): Likewise.
17583         * recog.c (peep2_find_free_register): Likewise.
17584         * ree.c (combine_reaching_defs): Likewise.
17585         * regcprop.c (maybe_mode_change): Likewise.
17586         * reginfo.c (init_reg_sets_1): Likewise.
17587         (choose_hard_reg_mode): Likewise.
17588         (simplifiable_subregs): Likewise.
17589         * regrename.c (check_new_reg_p): Likewise.
17590         * reload.c (find_valid_class): Likewise.
17591         (find_valid_class_1): Likewise.
17592         (reload_inner_reg_of_subreg): Likewise.
17593         (push_reload): Likewise.
17594         (combine_reloads): Likewise.
17595         (find_dummy_reload): Likewise.
17596         (find_reloads): Likewise.
17597         * reload1.c (find_reg): Likewise.
17598         (set_reload_reg): Likewise.
17599         (allocate_reload_reg): Likewise.
17600         (choose_reload_regs): Likewise.
17601         (reload_adjust_reg_for_temp): Likewise.
17602         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17603         (simplify_subreg_regno): Likewise.
17604         * sel-sched.c (init_regs_for_mode): Likewise.
17605         * varasm.c (make_decl_rtl): Likewise.
17606         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17607         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17608         HARD_REGNO_MODE_OK.
17609         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17610         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17611         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17612         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17613         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17614         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17615         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17616         (arc_mode_class): Delete.
17617         (HARD_REGNO_MODE_OK): Delete.
17618         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17619         (arc_hard_regno_mode_ok): Rename old array to...
17620         (arc_hard_regno_mode_ok_modes): ...this.
17621         (arc_conditional_register_usage): Update accordingly.
17622         (arc_mode_class): Make static.
17623         (arc_hard_regno_mode_ok): New function.
17624         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17625         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17626         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17627         (arm_hard_regno_mode_ok): Make static.
17628         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17629         HARD_REGNO_MODE_OK.
17630         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17631         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17632         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17633         return a bool.
17634         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17635         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17636         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17637         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17638         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17639         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17640         * config/bfin/predicates.md (valid_reg_operand): Use
17641         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17642         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17643         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17644         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17645         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17646         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17647         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17648         (cr16_hard_regno_mode_ok): Make static and return a bool.
17649         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17650         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17651         (cris_hard_regno_mode_ok): New function.
17652         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17653         (epiphany_mode_class): Delete.
17654         (HARD_REGNO_MODE_OK): Delete.
17655         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17656         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17657         (hard_regno_mode_ok): Rename to...
17658         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17659         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17660         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17661         HARD_REGNO_MODE_OK.
17662         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17663         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17664         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17665         (frv_hard_regno_mode_ok): Make static and return a bool.
17666         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17667         HARD_REGNO_MODE_OK.
17668         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17669         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17670         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17671         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17672         and return a bool.
17673         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17674         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17675         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17676         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17677         return a bool.
17678         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17679         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17680         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17681         (ia64_hard_regno_mode_ok): New function.
17682         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17683         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17684         (iq2000_hard_regno_mode_ok): New function.
17685         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17686         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17687         (lm32_hard_regno_mode_ok): New function.
17688         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17689         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17690         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17691         instead of HARD_REGNO_MODE_OK.
17692         (m32c_hard_regno_ok): Rename to...
17693         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17694         (m32c_cannot_change_mode_class): Update accordingly.
17695         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17696         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17697         (m32r_mode_class): Delete.
17698         (HARD_REGNO_MODE_OK): Delete.
17699         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17700         (m32r_hard_regno_mode_ok): Rename to...
17701         (m32r_hard_regno_modes): ...this.
17702         (m32r_mode_class): Make static.
17703         (m32r_hard_regno_mode_ok): New function.
17704         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17705         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17706         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17707         (m68k_hard_regno_mode_ok): Make static.
17708         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17709         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17710         (mcore_hard_regno_mode_ok): New function.
17711         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17712         (HARD_REGNO_MODE_OK): Delete.
17713         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17714         Rename to...
17715         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17716         (microblaze_hard_regno_mode_ok): New function.
17717         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17718         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17719         (mips_hard_regno_mode_ok): Delete.
17720         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17721         (mips_hard_regno_mode_ok_p): ...this and make static.
17722         (mips_hard_regno_mode_ok_p): Rename to...
17723         (mips_hard_regno_mode_ok_uncached): ...this.
17724         (mips_hard_regno_mode_ok): New function.
17725         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17726         of HARD_REGNO_MODE_OK.
17727         (mips_option_override): Update after above name changes.
17728         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17729         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17730         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17731         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17732         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17733         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17734         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17735         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17736         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17737         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17738         (msp430_hard_regno_mode_ok): Make static and return a bool.
17739         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17740         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17741         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17742         and return a bool.
17743         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17744         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17745         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17746         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17747         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17748         (PA_HARD_REGNO_MODE_OK): ...this
17749         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17750         (PA_HARD_REGNO_MODE_OK): ...this.
17751         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17752         (pa_hard_regno_mode_ok): New function.
17753         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17754         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17755         (pdp11_hard_regno_mode_ok): New function.
17756         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17757         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17758         Delete.
17759         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17760         Make static.
17761         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17762         (rs6000_hard_regno_mode_ok): Rename to...
17763         (rs6000_hard_regno_mode_ok_uncached): ...this.
17764         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17765         (rs6000_hard_regno_mode_ok): New function.
17766         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17767         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17768         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17769         (riscv_hard_regno_mode_ok): ...this and make static.
17770         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17771         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17772         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17773         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17774         (rl78_hard_regno_mode_ok): Make static and return bool.
17775         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17776         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17777         Delete.
17778         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17779         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17780         (rs6000_hard_regno_mode_ok): Rename to...
17781         (rs6000_hard_regno_mode_ok_uncached): ...this.
17782         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17783         (rs6000_hard_regno_mode_ok): New function.
17784         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17785         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17786         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17787         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17788         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17789         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17790         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17791         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17792         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17793         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17794         (sh_hard_regno_mode_ok): Make static.
17795         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17796         instead of HARD_REGNO_MODE_OK.
17797         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17798         (sparc_mode_class): Delete.
17799         (HARD_REGNO_MODE_OK): Delete.
17800         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17801         (hard_regno_mode_classes): Make static.
17802         (sparc_mode_class): Likewise.
17803         (sparc_hard_regno_mode_ok): New function.
17804         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17805         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17806         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17807         function.
17808         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17809         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17810         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17811         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17812         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17813         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17814         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17815         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17816         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17817         (visium_hard_regno_mode_ok): New function.
17818         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17819         instead of HARD_REGNO_MODE_OK.
17820         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17821         (HARD_REGNO_MODE_OK): Delete.
17822         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17823         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17824         (xtensa_option_override): Update accordingly.
17825         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17826         (xtensa_hard_regno_mode_ok): New function.
17827         * system.h (HARD_REGNO_MODE_OK): Poison.
17829 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17830             Alan Hayward  <alan.hayward@arm.com>
17831             David Sherwood  <david.sherwood@arm.com>
17833         * target.def (hard_regno_call_part_clobbered): New hook.
17834         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17835         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17836         * doc/tm.texi: Regenerate.
17837         * hooks.h (hook_bool_uint_mode_false): Declare.
17838         * hooks.c (hook_bool_uint_mode_false): New function.
17839         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17840         * cselib.c (cselib_process_insn): Use
17841         targetm.hard_regno_call_part_clobbered instead of
17842         HARD_REGNO_CALL_PART_CLOBBERED.
17843         * ira-conflicts.c (ira_build_conflicts): Likewise.
17844         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17845         * lra-constraints.c (need_for_call_save_p): Likewise.
17846         * lra-lives.c: Include target.h.
17847         (check_pseudos_live_through_calls): Use
17848         targetm.hard_regno_call_part_clobbered instead of
17849         HARD_REGNO_CALL_PART_CLOBBERED.
17850         * regcprop.c: Include target.h.
17851         (copyprop_hardreg_forward_1): Use
17852         targetm.hard_regno_call_part_clobbered instead of
17853         HARD_REGNO_CALL_PART_CLOBBERED.
17854         * reginfo.c (choose_hard_reg_mode): Likewise.
17855         * regrename.c (check_new_reg_p): Likewise.
17856         * reload.c (find_equiv_reg): Likewise.
17857         * reload1.c (emit_reload_insns): Likewise.
17858         * sched-deps.c (deps_analyze_insn): Likewise.
17859         * sel-sched.c (init_regs_for_mode): Likewise.
17860         (mark_unavailable_hard_regs): Likewise.
17861         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17862         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17863         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
17864         New function.
17865         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17866         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17867         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
17868         Delete.
17869         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
17870         and return a bool.
17871         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17872         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17873         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
17874         function.
17875         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17876         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17877         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
17878         function.
17879         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17880         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
17881         Delete.
17882         * config/powerpcspe/powerpcspe.c
17883         (rs6000_hard_regno_call_part_clobbered): New function.
17884         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17885         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17886         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
17887         New function.
17888         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17889         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17890         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
17891         function.
17892         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17893         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17894         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
17896 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17897             Alan Hayward  <alan.hayward@arm.com>
17898             David Sherwood  <david.sherwood@arm.com>
17900         * rtl.h (subreg_memory_offset): Declare.
17901         * emit-rtl.c (subreg_memory_offset): New function.
17902         * expmed.c (store_bit_field_1): Use it.
17903         * expr.c (undefined_operand_subword_p): Likewise.
17904         * simplify-rtx.c (simplify_subreg): Likewise.
17906 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
17908         PR rtl-optimization/57448
17909         PR target/67458
17910         PR target/81316
17911         * optabs.c (expand_atomic_load): Place compiler memory barriers if
17912         using atomic_load pattern.
17913         (expand_atomic_store): Likewise.
17915 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
17917         PR sanitizer/81981
17918         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
17919         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
17920         handling.  Use replace_call_with_value with NULL instead of
17921         gsi_replace, unlink_stmt_vdef and release_defs.
17923         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
17924         instead of tab.
17926         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
17928 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17930         PR bootstrap/82045
17931         * rtl.h (emit_library_call_value_1): Declare.
17932         (emit_library_call): Replace declaration with a series of overloads.
17933         Remove the parameter count argument.
17934         (emit_library_call_value): Likewise.
17935         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
17936         with an "rtx_mode_t *".
17937         (emit_library_call_value): Delete.
17938         (emit_library_call): Likewise.
17939         * asan.c (asan_emit_stack_protection): Update calls accordingly.
17940         (asan_emit_allocas_unpoison): Likewise.
17941         * builtins.c (expand_builtin_powi): Likewise.
17942         (expand_asan_emit_allocas_unpoison): Likewise.
17943         * cfgexpand.c (expand_main_function): Likewise.
17944         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
17945         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
17946         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
17947         * config/arm/arm.c (arm_trampoline_init): Likewise.
17948         (arm_call_tls_get_addr): Likewise.
17949         (arm_expand_divmod_libfunc): Likewise.
17950         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
17951         (smulsi3_highpart): Likewise.
17952         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
17953         (c6x_expand_compare): Likewise.
17954         (c6x_expand_movmem): Likewise.
17955         * config/frv/frv.c (frv_trampoline_init): Likewise.
17956         * config/i386/i386.c (ix86_trampoline_init): Likewise.
17957         (ix86_expand_divmod_libfunc): Likewise.
17958         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
17959         (ia64_expand_compare): Likewise.
17960         (ia64_profile_hook): Likewise.
17961         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
17962         (nonlocal_goto): Likewise.
17963         (restore_stack_nonlocal): Likewise.
17964         * config/m32r/m32r.c (block_move_call): Likewise.
17965         (m32r_trampoline_init): Likewise.
17966         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
17967         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
17968         (m68k_call_m68k_read_tp): Likewise.
17969         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
17970         (microblaze_expand_divide): Likewise.
17971         * config/mips/mips.h (mips_args): Likewise.
17972         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
17973         (MIPS_ICACHE_SYNC): Likewise.
17974         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
17975         (nios2_trampoline_init): Likewise.
17976         * config/pa/pa.c (hppa_tls_call): Likewise.
17977         (pa_trampoline_init): Likewise.
17978         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
17979         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
17980         (expand_strn_compare): Likewise.
17981         (rs6000_generate_compare): Likewise.
17982         (rs6000_expand_float128_convert): Likewise.
17983         (output_profile_hook): Likewise.
17984         (rs6000_trampoline_init): Likewise.
17985         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
17986         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
17987         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
17988         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
17989         (rs6000_generate_compare): Likewise.
17990         (rs6000_expand_float128_convert): Likewise.
17991         (output_profile_hook): Likewise.
17992         (rs6000_trampoline_init): Likewise.
17993         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
17994         * config/sh/sh.c (sh_trampoline_init): Likewise.
17995         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
17996         (sparc_emit_float_lib_cmp): Likewise.
17997         (sparc32_initialize_trampoline): Likewise.
17998         (sparc64_initialize_trampoline): Likewise.
17999         (sparc_profile_hook): Likewise.
18000         * config/spu/spu.c (ea_load_store): Likewise.
18001         * config/spu/spu.md (floatunssidf2): Likewise.
18002         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18003         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18004         * config/visium/visium.c (expand_block_move_4): Likewise.
18005         (expand_block_move_2): Likewise.
18006         (expand_block_move_1): Likewise.
18007         (expand_block_set_4): Likewise.
18008         (expand_block_set_2): Likewise.
18009         (expand_block_set_1): Likewise.
18010         (visium_trampoline_init): Likewise.
18011         (visium_profile_hook): Likewise.
18012         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18013         (xtensa_setup_frame_addresses): Likewise.
18014         (xtensa_trampoline_init): Likewise.
18015         * except.c (sjlj_emit_function_enter): Likewise.
18016         (sjlj_emit_function_exit): Likewise.
18017         * explow.c (allocate_dynamic_stack_space): Likewise.
18018         (probe_stack_range): Likewise.
18019         * expr.c (convert_mode_scalar): Likewise.
18020         * optabs.c (expand_binop): Likewise.
18021         (expand_twoval_binop_libfunc): Likewise.
18022         (expand_unop): Likewise.
18023         (prepare_cmp_insn): Likewise.
18024         (prepare_float_lib_cmp): Likewise.
18025         (expand_float): Likewise.
18026         (expand_fix): Likewise.
18027         (expand_fixed_convert): Likewise.
18028         (maybe_emit_sync_lock_test_and_set): Likewise.
18029         (expand_atomic_compare_and_swap): Likewise.
18030         (expand_mem_thread_fence): Likewise.
18031         (expand_atomic_fetch_op): Likewise.
18033 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18035         * doc/generic.texi (OpenACC): Adjust URL.
18036         * doc/invoke.texi (C Dialect Options): Ditto.
18038 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18040         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18041         predicate for operand 1.  Add (m,<S>) constraint.
18042         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18043         Prevent memory operand 1 with register operand 2.
18045 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18047         PR rtl-optimization/82024
18048         * combine.c (try_combine): If the combination result is a PARALLEL,
18049         and we only need to retain the SET in there that would be placed
18050         at I2, check that we can place that at I3 instead, before doing so.
18052 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18054         PR target/81766
18055         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18056         instead of void.
18057         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18058         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18059         and label.
18061 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18062             Jeff Law  <law@redhat.com>
18064         * varasm.c (bss_initializer_p): Do not put constants into .bss
18065         (categorize_decl_for_section): Handle bss_initializer_p returning
18066         false when DECL_INITIAL is NULL.
18068 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18070         PR target/82012
18071         * config/s390/s390.c (s390_can_inline_p): New function.
18073 2017-09-01  Jeff Law  <law@redhat.com>
18075         PR tree-optimization/82052
18076         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18077         Always initialize the returned slot after a hash table miss
18078         when INSERT is true.
18080 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18082         * config/s390/s390.md (mem_signal_fence): Remove.
18083         * doc/md.texi (mem_signal_fence): Remove.
18084         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18085         Update comments.
18086         * target-insns.def (mem_signal_fence): Remove.
18088 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18090         PR sanitizer/81902
18091         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18093         PR sanitizer/81923
18094         * asan.c (create_odr_indicator): Strip name encoding from assembler
18095         name before appending it after __odr_asan_.
18097 2017-09-01  Martin Liska  <mliska@suse.cz>
18099         PR tree-optimization/82059
18100         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18101         frequency only when an edge is redirected.
18103 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18105         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18106         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18107         (arc_conditional_register_usage): Remove ARC600 lp_count
18108         exception.
18109         (arc_file_start): Emit Tag_ARC_CPU_variation.
18110         (arc_can_use_doloop_p): New conditions to use ZOLs.
18111         (hwloop_fail): New function.
18112         (hwloop_optimize): Likewise.
18113         (hwloop_pattern_reg): Likewise.
18114         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18115         (arc_reorg_loops): New function, calls reorg_loops.
18116         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18117         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18118         hwloop_optimize.
18119         (arc_loop_hazard): Remove function, functionality moved into
18120         hwloop_optimize.
18121         (arc_hazard): Remove arc_loop_hazard call.
18122         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18123         into hwloop_optimize.
18124         (arc_label_align): Remove ZOL handling.
18125         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18126         * config/arc/arc.md (doloop_begin): Remove pattern.
18127         (doloop_begin_i): Likewise.
18128         (doloop_end_i): Likewise.
18129         (doloop_fallback): Likewise.
18130         (doloop_fallback_m): Likewise.
18131         (doloop_end): Reimplement expand.
18132         (arc_lp): New pattern for LP instruction.
18133         (loop_end): New pattern.
18134         (loop_fail): Likewise.
18135         (decrement_and_branch_until_zero): Likewise.
18136         * config/arc/arc.opt (mlpc-width): New option.
18137         * doc/invoke.texi (mlpc-width): Document option.
18139 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18141         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18142         (arc_ccfsm_advance): Fix checking for delay slots.
18143         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18145 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18147         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18148         immediates.
18149         (movhi_insn): Update store instruction constraint which are saving
18150         6-bit short immediates.
18151         (movsi_insn): Consider also short scaled load operations.
18152         (zero_extendhisi2_i): Use Usd constraint instead of T.
18153         (extendhisi2_i): Add q constraint.
18154         (arc_clzsi2): Add type and length attributes.
18155         (arc_ctzsi2): Likewise.
18156         * config/arc/constraints.md (Usc): Update constraint, the
18157         assembler can parse two relocations for a single instruction.
18159 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18161         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18162         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18164 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18166         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18167         match as powerpc-wrs-vxworks*.
18169 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18171         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18172         register constraint for by-element operand.
18173         (aarch64_mls_elt_merge<mode>): Likewise.
18175 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18177         * config/arc/arc.c (arc_can_follow_jump): Check for short
18178         branches.
18180 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18182         * config.gcc: Use g.opt for arc.
18183         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18184         functionality moved to ...
18185         (legitimate_scaled_address_p): New function, ...here.
18186         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18187         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18188         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18189         condition.
18190         (arc_override_options): Handle G option.
18191         (arc_output_pic_addr_const): Correct function definition.
18192         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18193         (arc_decl_anon_ns_mem_p): Delete.
18194         (arc_in_small_data_p): Overhaul this function to take into
18195         consideration the value given via G option.
18196         (arc_rewrite_small_data_1): Renamed and corrected old
18197         arc_rewrite_small_data function.
18198         (arc_rewrite_small_data): New function.
18199         (small_data_pattern): Don't use pic_offset_table_rtx.
18200         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18201         * config/arc/simdext.md (movmisalignv2hi): Use
18202         prepare_move_operands function.
18203         (mov*): Likewise.
18204         (movmisalign*): Likewise.
18205         * doc/invoke.texi (ARC options): Document -G option.
18207 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18209         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18210         prototype.
18211         * config/arc/arc.c (arc_print_operand): Output scalled address for
18212         sdata whenever is possible.
18213         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18214         load/stores are available.
18215         (compact_sda_memory_operand): Check for the alignment required by
18216         code density instructions.
18217         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18218         constraint.
18219         * config/arc/constraints.md (Usd): Update constraint.
18220         (Us0): New constraint.
18221         (Usc): Update constraint.
18223 2017-08-31  Richard Biener  <rguenther@suse.de>
18225         PR middle-end/82054
18226         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18227         function only once.
18229 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18231         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18232         Resize type_signature.
18234 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18235             Alan Hayward  <alan.hayward@arm.com>
18236             David Sherwood  <david.sherwood@arm.com>
18238         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18239         subregs whose inner modes can be stored in GPRs.
18240         (aarch64_classify_index): Likewise.
18242 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18243             Alan Hayward  <alan.hayward@arm.com>
18244             David Sherwood  <david.sherwood@arm.com>
18246         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18247         (V_INT_EQUIV): ...this.
18248         (v_cmp_result): Rename to...
18249         (v_int_equiv): ...this.
18250         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18251         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18252         (copysign<mode>3): Likewise.
18253         (aarch64_simd_bsl<mode>_internal): Likewise.
18254         (aarch64_simd_bsl<mode>): Likewise.
18255         (vec_cmp<mode><mode>): Likewise.
18256         (vcond<mode><mode>): Likewise.
18257         (vcond<v_cmp_mixed><mode>): Likewise.
18258         (vcondu<mode><v_cmp_mixed>): Likewise.
18259         (aarch64_cm<optab><mode>): Likewise.
18260         (aarch64_cmtst<mode>): Likewise.
18261         (aarch64_fac<optab><mode>): Likewise.
18262         (vec_perm_const<mode>): Likewise.
18263         (vcond_mask_<mode><v_cmp_result>): Rename to...
18264         (vcond_mask_<mode><v_int_equiv>): ...this.
18265         (vec_cmp<mode><v_cmp_result>): Rename to...
18266         (vec_cmp<mode><v_int_equiv>): ...this.
18268 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18269             Alan Hayward  <alan.hayward@arm.com>
18270             David Sherwood  <david.sherwood@arm.com>
18272         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18273         vector modes.
18274         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18275         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18276         (UNSPEC_LD4_DREG): New unspecs.
18277         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18278         (aarch64_ld2<mode>_dreg_be): Replace with...
18279         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18280         unspec.
18281         (aarch64_ld3<mode>_dreg_le)
18282         (aarch64_ld3<mode>_dreg_be): Replace with...
18283         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18284         unspec.
18285         (aarch64_ld4<mode>_dreg_le)
18286         (aarch64_ld4<mode>_dreg_be): Replace with...
18287         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18288         unspec.
18290 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18292         PR tree-optimization/81987
18293         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18294         insert an initializer in a location not dominated by the stride
18295         definition.
18297 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18299         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18300         on the entire header of the finally block in the fallthru case.
18302 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18304         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18306 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18308         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18309         rs6000_emit_move_from_cr and call renamed function.
18310         (rs6000_emit_prologue): Call renamed functions.
18311         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18312         movesi_from_cr, remove volatile CRs.
18314 2017-08-30  Jon Beniston  <jon@beniston.com>
18315             Richard Biener  <rguenther@suse.de>
18317         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18318         of VECTOR_MODE_P check.
18319         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18320         element vector types.
18322 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18324         * df.h (df_read_modify_subreg_p): Remove in favor of...
18325         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18326         const_rtx instead of an rtx.
18327         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18328         * df-problems.c (df_word_lr_mark_ref): Likewise.
18329         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18330         (mark_pseudo_reg_dead): Likewise.
18331         (mark_ref_dead): Likewise.
18332         * reginfo.c (init_subregs_of_mode): Likewise.
18333         * sched-deps.c (sched_analyze_1): Likewise.
18334         * df-scan.c (df_def_record_1): Likewise.
18335         (df_uses_record): Likewise.
18336         (df_read_modify_subreg_p): Remove in favor of...
18337         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18338         const_rtx instead of an rtx.
18340 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18341             Alan Hayward  <alan.hayward@arm.com>
18342             David Sherwood  <david.sherwood@arm.com>
18344         * rtl.h (partial_subreg_p): New function.
18345         * caller-save.c (save_call_clobbered_regs): Use it.
18346         * calls.c (expand_call): Likewise.
18347         * combine.c (combinable_i3pat): Likewise.
18348         (simplify_set): Likewise.
18349         (make_extraction): Likewise.
18350         (make_compound_operation_int): Likewise.
18351         (gen_lowpart_or_truncate): Likewise.
18352         (force_to_mode): Likewise.
18353         (make_field_assignment): Likewise.
18354         (reg_truncated_to_mode): Likewise.
18355         (record_truncated_value): Likewise.
18356         (move_deaths): Likewise.
18357         * cse.c (record_jump_cond): Likewise.
18358         (cse_insn): Likewise.
18359         * cselib.c (cselib_lookup_1): Likewise.
18360         * expmed.c (extract_bit_field_using_extv): Likewise.
18361         * function.c (assign_parm_setup_reg): Likewise.
18362         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18363         * ira-build.c (create_insn_allocnos): Likewise.
18364         * lra-coalesce.c (merge_pseudos): Likewise.
18365         * lra-constraints.c (match_reload): Likewise.
18366         (simplify_operand_subreg): Likewise.
18367         (curr_insn_transform): Likewise.
18368         * lra-lives.c (process_bb_lives): Likewise.
18369         * lra.c (new_insn_reg): Likewise.
18370         (lra_substitute_pseudo): Likewise.
18371         * regcprop.c (mode_change_ok): Likewise.
18372         (maybe_mode_change): Likewise.
18373         (copyprop_hardreg_forward_1): Likewise.
18374         * reload.c (push_reload): Likewise.
18375         (find_reloads): Likewise.
18376         (find_reloads_subreg_address): Likewise.
18377         * reload1.c (alter_reg): Likewise.
18378         (eliminate_regs_1): Likewise.
18379         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18381 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18383         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18384         back to if statements, including unpack.
18386 2017-08-30  Martin Liska  <mliska@suse.cz>
18388         PR inline-asm/82001
18389         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18390         Rename to ...
18391         (func_checker::compare_asm_inputs_outputs): ... this function.
18392         (func_checker::compare_gimple_asm): Use the function to compare
18393         also ASM constrains.
18394         * ipa-icf-gimple.h: Rename the function.
18396 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18397             Alan Hayward  <alan.hayward@arm.com>
18398             David Sherwood  <david.sherwood@arm.com>
18400         * coretypes.h (complex_mode): New type.
18401         * gdbhooks.py (build_pretty_printer): Handle it.
18402         * machmode.h (complex_mode): New class.
18403         (complex_mode::includes_p): New function.
18404         (is_complex_int_mode): Likewise.
18405         (is_complex_float_mode): Likewise.
18406         * genmodes.c (get_mode_class): Handle complex mode classes.
18407         * function.c (expand_function_end): Use is_complex_int_mode.
18409 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18410             Alan Hayward  <alan.hayward@arm.com>
18411             David Sherwood  <david.sherwood@arm.com>
18413         * coretypes.h (scalar_mode_pod): New typedef.
18414         * gdbhooks.py (build_pretty_printer): Handle it.
18415         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18416         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18417         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18418         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18419         as_a <scalar_mode>.
18421 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18422             Alan Hayward  <alan.hayward@arm.com>
18423             David Sherwood  <david.sherwood@arm.com>
18425         * machmode.h (mode_for_vector): Take a scalar_mode instead
18426         of a machine_mode.
18427         * stor-layout.c (mode_for_vector): Likewise.
18428         * explow.c (promote_mode): Use as_a <scalar_mode>.
18429         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18431 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18432             Alan Hayward  <alan.hayward@arm.com>
18433             David Sherwood  <david.sherwood@arm.com>
18435         * target.def (preferred_simd_mode): Take a scalar_mode
18436         instead of a machine_mode.
18437         * targhooks.h (default_preferred_simd_mode): Likewise.
18438         * targhooks.c (default_preferred_simd_mode): Likewise.
18439         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18440         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18441         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18442         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18443         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18444         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18445         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18446         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18447         Likewise.
18448         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18449         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18450         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18451         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18452         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18453         * doc/tm.texi: Regenerate.
18454         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18455         non-scalar modes.
18457 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18458             Alan Hayward  <alan.hayward@arm.com>
18459             David Sherwood  <david.sherwood@arm.com>
18461         * target.def (scalar_mode_supported_p): Take a scalar_mode
18462         instead of a machine_mode.
18463         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18464         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18465         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18466         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18467         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18468         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18469         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18470         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18471         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18472         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18473         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18474         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18475         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18476         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18477         Likewise.
18478         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18479         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18480         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18481         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18482         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18483         Likewise.
18484         * doc/tm.texi: Regenerate.
18486 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18487             Alan Hayward  <alan.hayward@arm.com>
18488             David Sherwood  <david.sherwood@arm.com>
18490         * coretypes.h (opt_scalar_mode): New typedef.
18491         * gdbhooks.py (build_pretty_printers): Handle it.
18492         * machmode.h (mode_iterator::get_2xwider): Add overload for
18493         opt_mode<T>.
18494         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18495         over scalar modes.
18496         * expr.c (convert_mode_scalar): Likewise.
18497         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18498         * optabs.c (expand_float): Likewise.
18499         (expand_fix): Likewise.
18500         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18502 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18503             Alan Hayward  <alan.hayward@arm.com>
18504             David Sherwood  <david.sherwood@arm.com>
18506         * optabs.c (expand_float): Explicitly check for scalars before
18507         using a branching expansion.
18508         (expand_fix): Likewise.
18510 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18511             Alan Hayward  <alan.hayward@arm.com>
18512             David Sherwood  <david.sherwood@arm.com>
18514         * expr.c (convert_mode): Split scalar handling out into...
18515         (convert_mode_scalar): ...this new function.  Treat the modes
18516         as scalar_modes.
18518 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18519             Alan Hayward  <alan.hayward@arm.com>
18520             David Sherwood  <david.sherwood@arm.com>
18522         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18523         and scalar_mode.
18524         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18526 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18527             Alan Hayward  <alan.hayward@arm.com>
18528             David Sherwood  <david.sherwood@arm.com>
18530         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18531         rather than a machine_mode.
18532         (fixed_from_string): Likewise.
18533         (fixed_convert): Likewise.
18534         (fixed_convert_from_int): Likewise.
18535         (fixed_convert_from_real): Likewise.
18536         (real_convert_from_fixed): Likewise.
18537         * fixed-value.c (fixed_from_double_int): Likewise.
18538         (fixed_from_string): Likewise.
18539         (fixed_convert): Likewise.
18540         (fixed_convert_from_int): Likewise.
18541         (fixed_convert_from_real): Likewise.
18542         (real_convert_from_fixed): Likewise.
18543         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18545 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18546             Alan Hayward  <alan.hayward@arm.com>
18547             David Sherwood  <david.sherwood@arm.com>
18549         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18550         of separate mode class checks.  Do not allow vector modes here.
18551         (immed_wide_int_const): Use as_a <scalar_mode>.
18552         * explow.c (trunc_int_for_mode): Likewise.
18553         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18554         (wi::shwi): Likewise.
18555         (wi::min_value): Likewise.
18556         (wi::max_value): Likewise.
18557         * dwarf2out.c (loc_descriptor): Likewise.
18558         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18559         for CONST_WIDE_INT.
18561 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18562             Alan Hayward  <alan.hayward@arm.com>
18563             David Sherwood  <david.sherwood@arm.com>
18565         * tree.h (SCALAR_TYPE_MODE): New macro.
18566         * expr.c (expand_expr_addr_expr_1): Use it.
18567         (expand_expr_real_2): Likewise.
18568         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18569         (fold_convert_const_fixed_from_int): Likewise.
18570         (fold_convert_const_fixed_from_real): Likewise.
18571         (native_encode_fixed): Likewise
18572         (native_encode_complex): Likewise
18573         (native_encode_vector): Likewise.
18574         (native_interpret_fixed): Likewise.
18575         (native_interpret_real): Likewise.
18576         (native_interpret_complex): Likewise.
18577         (native_interpret_vector): Likewise.
18578         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18579         (simd_clone_adjust_argument_types): Likewise.
18580         (simd_clone_init_simd_arrays): Likewise.
18581         (simd_clone_adjust): Likewise.
18582         * stor-layout.c (layout_type): Likewise.
18583         * tree.c (build_minus_one_cst): Likewise.
18584         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18585         * tree-inline.c (estimate_move_cost): Likewise.
18586         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18587         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18588         (vectorizable_reduction): Likewise.
18589         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18590         (vect_recog_mixed_size_cond_pattern): Likewise.
18591         (check_bool_pattern): Likewise.
18592         (adjust_bool_pattern): Likewise.
18593         (search_type_for_mask_1): Likewise.
18594         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18595         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18596         (vectorizable_load): Likewise.
18597         (vectorizable_store): Likewise.
18598         * ubsan.c (ubsan_encode_value): Likewise.
18599         * varasm.c (output_constant): Likewise.
18601 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18602             Alan Hayward  <alan.hayward@arm.com>
18603             David Sherwood  <david.sherwood@arm.com>
18605         * coretypes.h (scalar_mode): New class.
18606         * machmode.h (scalar_mode): Likewise.
18607         (scalar_mode::includes_p): New function.
18608         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18609         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18610         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18611         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18612         * expmed.c (store_bit_field_1): Likewise.
18613         (extract_bit_field_1): Likewise.
18614         * expr.c (write_complex_part): Likewise.
18615         (read_complex_part): Likewise.
18616         (emit_move_complex_push): Likewise.
18617         (expand_expr_real_2): Likewise.
18618         * function.c (assign_parm_setup_reg): Likewise.
18619         (assign_parms_unsplit_complex): Likewise.
18620         * optabs.c (expand_binop): Likewise.
18621         * rtlanal.c (subreg_get_info): Likewise.
18622         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18623         * varasm.c (output_constant_pool_2): Likewise.
18625 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18626             Alan Hayward  <alan.hayward@arm.com>
18627             David Sherwood  <david.sherwood@arm.com>
18629         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18630         assertion.
18631         (expmed_mult_highpart_optab): Likewise.
18632         (expmed_mult_highpart): Likewise.
18634 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18635             Alan Hayward  <alan.hayward@arm.com>
18636             David Sherwood  <david.sherwood@arm.com>
18638         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18639         instead of a machine_mode.
18640         (builtin_memset_read_str): Likewise.
18641         * builtins.c (c_readstr): Likewise.
18642         (builtin_memcpy_read_str): Likewise.
18643         (builtin_strncpy_read_str): Likewise.
18644         (builtin_memset_read_str): Likewise.
18645         (builtin_memset_gen_str): Likewise.
18646         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18647         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18648         instead of a machine_mode.
18649         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18650         variables.
18651         (make_extraction): Likewise.
18652         (try_widen_shift_mode): Take and return scalar_int_modes instead
18653         of machine_modes.
18654         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18655         a scalar_int_mode instead of a machine_mode.
18656         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18657         (avr_addr_space_pointer_mode): Likewise.
18658         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18659         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18660         (msp430_unwind_word_mode): Likewise.
18661         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18662         (spu_addr_space_pointer_mode): Likewise.
18663         (spu_addr_space_address_mode): Likewise.
18664         (spu_libgcc_cmp_return_mode): Likewise.
18665         (spu_libgcc_shift_count_mode): Likewise.
18666         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18667         (rl78_addr_space_pointer_mode): Likewise.
18668         (fl78_unwind_word_mode): Likewise.
18669         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18670         machine_mode.
18671         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18672         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18673         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18674         (mips_valid_pointer_mode): Likewise.
18675         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18676         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18677         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18678         of a machine_mode.
18679         (ft32_addr_space_address_mode): Likewise.
18680         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18681         scalar_int_mode instead of a machine_mode.
18682         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18683         of a machine_mode.
18684         (m32c_addr_space_address_mode): Likewise.
18685         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18686         (rs6000_eh_return_filter_mode): Likewise.
18687         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18688         (rs6000_eh_return_filter_mode): Likewise.
18689         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18690         (s390_libgcc_shift_count_mode): Likewise.
18691         (s390_unwind_word_mode): Likewise.
18692         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18693         machine_mode.
18694         * target.def (mode_rep_extended): Likewise.
18695         (valid_pointer_mode): Likewise.
18696         (addr_space.valid_pointer_mode): Likewise.
18697         (eh_return_filter_mode): Return a scalar_int_mode rather than
18698         a machine_mode.
18699         (libgcc_cmp_return_mode): Likewise.
18700         (libgcc_shift_count_mode): Likewise.
18701         (unwind_word_mode): Likewise.
18702         (addr_space.pointer_mode): Likewise.
18703         (addr_space.address_mode): Likewise.
18704         * doc/tm.texi: Regenerate.
18705         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18706         a machine_mode.
18707         (do_jump): Use scalar_int_mode for local variables.
18708         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18709         rather than a machine_mode.
18710         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18711         (scompare_loc_descriptor_wide): Likewise.
18712         (scompare_loc_descriptor_narrow): Likewise.
18713         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18714         variables.
18715         * except.c (sjlj_emit_dispatch_table): Likewise.
18716         (expand_builtin_eh_copy_values): Likewise.
18717         * explow.c (convert_memory_address_addr_space_1): Likewise.
18718         Take a scalar_int_mode rather than a machine_mode.
18719         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18720         than a machine_mode.
18721         (memory_address_addr_space): Use scalar_int_mode for local variables.
18722         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18723         rather than a machine_mode.
18724         * expmed.c (mask_rtx): Likewise.
18725         (init_expmed_one_conv): Likewise.
18726         (expand_mult_highpart_adjust): Likewise.
18727         (extract_high_half): Likewise.
18728         (expmed_mult_highpart_optab): Likewise.
18729         (expmed_mult_highpart): Likewise.
18730         (expand_smod_pow2): Likewise.
18731         (expand_sdiv_pow2): Likewise.
18732         (emit_store_flag_int): Likewise.
18733         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18734         variables.
18735         (extract_low_bits): Likewise.
18736         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18737         a machine_mode.
18738         * expr.c (pieces_addr::adjust):  Likewise.
18739         (can_store_by_pieces): Likewise.
18740         (store_by_pieces): Likewise.
18741         (clear_by_pieces_1): Likewise.
18742         (expand_expr_addr_expr_1): Likewise.
18743         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18744         (expand_expr_real_1): Likewise.
18745         (try_casesi): Likewise.
18746         * final.c (shorten_branches): Likewise.
18747         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18748         type of "mode" to machine_mode.
18749         * internal-fn.c (expand_arith_overflow_result_store): Take a
18750         scalar_int_mode rather than a machine_mode.
18751         (expand_mul_overflow): Use scalar_int_mode for local variables.
18752         * loop-doloop.c (doloop_modify): Likewise.
18753         (doloop_optimize): Likewise.
18754         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18755         than a machine_mode.
18756         (expand_doubleword_shift_condmove): Likewise.
18757         (expand_doubleword_shift): Likewise.
18758         (expand_doubleword_clz): Likewise.
18759         (expand_doubleword_popcount): Likewise.
18760         (expand_doubleword_parity): Likewise.
18761         (expand_absneg_bit): Use scalar_int_mode for local variables.
18762         (prepare_float_lib_cmp): Likewise.
18763         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18764         rather than a machine_mode.
18765         (convert_memory_address_addr_space): Likewise.
18766         (get_mode_bounds): Likewise.
18767         (get_address_mode): Return a scalar_int_mode rather than a
18768         machine_mode.
18769         * rtlanal.c (get_address_mode): Likewise.
18770         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18771         than a machine_mode.
18772         * targhooks.c (default_mode_rep_extended): Likewise.
18773         (default_valid_pointer_mode): Likewise.
18774         (default_addr_space_valid_pointer_mode): Likewise.
18775         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18776         than a machine_mode.
18777         (default_libgcc_cmp_return_mode): Likewise.
18778         (default_libgcc_shift_count_mode): Likewise.
18779         (default_unwind_word_mode): Likewise.
18780         (default_addr_space_pointer_mode): Likewise.
18781         (default_addr_space_address_mode): Likewise.
18782         * targhooks.h (default_eh_return_filter_mode): Likewise.
18783         (default_libgcc_cmp_return_mode): Likewise.
18784         (default_libgcc_shift_count_mode): Likewise.
18785         (default_unwind_word_mode): Likewise.
18786         (default_addr_space_pointer_mode): Likewise.
18787         (default_addr_space_address_mode): Likewise.
18788         (default_mode_rep_extended): Take a scalar_int_mode rather than
18789         a machine_mode.
18790         (default_valid_pointer_mode): Likewise.
18791         (default_addr_space_valid_pointer_mode): Likewise.
18792         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18793         local variables.
18794         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18795         rather than a machine_mode.
18796         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18797         for local variables.
18798         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18799         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18800         than a machine_mode.
18802 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18803             Alan Hayward  <alan.hayward@arm.com>
18804             David Sherwood  <david.sherwood@arm.com>
18806         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18807         the mode argument to scalar_int_mode.
18808         (do_jump_by_parts_zero_rtx): Likewise.
18809         (do_jump_by_parts_equality_rtx): Likewise.
18810         (do_jump_by_parts_greater): Take a mode argument.
18811         (do_jump_by_parts_equality): Likewise.
18812         (do_jump_1): Update calls accordingly.
18814 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18815             Alan Hayward  <alan.hayward@arm.com>
18816             David Sherwood  <david.sherwood@arm.com>
18818         * is-a.h (safe_dyn_cast): New function.
18819         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18820         (jump_table_for_label): Likewise.
18821         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18822         instead of an rtx_insn *.
18823         (shorten_branches): Use dyn_cast instead of LABEL_P and
18824         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18825         rtx_jump_table_data::get_data_mode.
18826         (final_scan_insn): Likewise.
18828 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18829             Alan Hayward  <alan.hayward@arm.com>
18830             David Sherwood  <david.sherwood@arm.com>
18832         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18833         trying to combine a full-register integer set with a subreg
18834         integer set.
18836 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18837             Alan Hayward  <alan.hayward@arm.com>
18838             David Sherwood  <david.sherwood@arm.com>
18840         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18841         that is always either address_mode or pointer_mode.
18843 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18844             Alan Hayward  <alan.hayward@arm.com>
18845             David Sherwood  <david.sherwood@arm.com>
18847         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18848         when the two are known to be equal.
18850 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18851             Alan Hayward  <alan.hayward@arm.com>
18852             David Sherwood  <david.sherwood@arm.com>
18854         * simplify-rtx.c (simplify_const_unary_operation): Use
18855         is_a <scalar_int_mode> instead of checking for a nonzero
18856         precision.  Forcibly convert op_mode to a scalar_int_mode
18857         in that case.  More clearly differentiate the operand and
18858         result modes and use the former when deciding what the value
18859         of a count-bits operation should be.  Use is_int_mode instead
18860         of checking for a MODE_INT.  Remove redundant check for whether
18861         this mode has a zero precision.
18863 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18864             Alan Hayward  <alan.hayward@arm.com>
18865             David Sherwood  <david.sherwood@arm.com>
18867         * optabs.c (widen_leading): Change the type of the mode argument
18868         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
18869         (widen_bswap): Likewise.
18870         (expand_parity): Likewise.
18871         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
18872         (expand_ffs): Likewise.
18873         (epand_unop): Check for scalar integer modes before calling the
18874         above routines.
18876 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18877             Alan Hayward  <alan.hayward@arm.com>
18878             David Sherwood  <david.sherwood@arm.com>
18880         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
18881         Expand commentary.
18882         (expand_expr_real_1): Update call accordingly.
18884 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18885             Alan Hayward  <alan.hayward@arm.com>
18886             David Sherwood  <david.sherwood@arm.com>
18888         * expmed.c (store_bit_field_using_insv): Add op0_mode and
18889         value_mode arguments.  Use scalar_int_mode internally.
18890         (store_bit_field_1): Rename the new integer mode from imode
18891         to op0_mode and use it instead of GET_MODE (op0).  Update calls
18892         to store_split_bit_field, store_bit_field_using_insv and
18893         store_fixed_bit_field.
18894         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
18895         Use scalar_int_mode internally.  Use a bit count rather than a mode
18896         when calculating the largest bit size for get_best_mode.
18897         Update calls to store_split_bit_field and store_fixed_bit_field_1.
18898         (store_fixed_bit_field_1): Add mode and value_mode arguments.
18899         Remove assertion that OP0 has a scalar integer mode.
18900         (store_split_bit_field): Add op0_mode and value_mode arguments.
18901         Update calls to extract_fixed_bit_field.
18902         (extract_bit_field_using_extv): Add an op0_mode argument.
18903         Use scalar_int_mode internally.
18904         (extract_bit_field_1): Rename the new integer mode from imode to
18905         op0_mode and use it instead of GET_MODE (op0).  Update calls to
18906         extract_split_bit_field, extract_bit_field_using_extv and
18907         extract_fixed_bit_field.
18908         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
18909         to extract_split_bit_field and extract_fixed_bit_field_1.
18910         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
18911         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
18912         on the target mode.
18913         (extract_split_bit_field): Add an op0_mode argument.  Update call
18914         to extract_fixed_bit_field.
18916 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18917             Alan Hayward  <alan.hayward@arm.com>
18918             David Sherwood  <david.sherwood@arm.com>
18920         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
18921         * explow.c (hard_function_value): Likewise.
18922         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
18923         convert_to_mode call outside the loop.
18924         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
18925         for the mode iterator.  Require the mode specified by max_pieces
18926         to exist.
18927         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
18928         mode iterator.
18929         (copy_blkmode_to_reg): Likewise.
18930         (set_storage_via_setmem): Likewise.
18931         * optabs.c (prepare_cmp_insn): Likewise.
18932         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
18933         * stor-layout.c (finish_bitfield_representative): Likewise.
18935 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18936             Alan Hayward  <alan.hayward@arm.com>
18937             David Sherwood  <david.sherwood@arm.com>
18939         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
18940         * expr.c (convert_move): Use them.
18941         (convert_modes): Likewise.
18942         (store_expr_with_bounds): Likewise.
18944 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18945             Alan Hayward  <alan.hayward@arm.com>
18946             David Sherwood  <david.sherwood@arm.com>
18948         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
18949         parameter for the mode of "x".  Remove the "known_x", "known_mode"
18950         and "known_ret" arguments.  Change the type of the mode argument
18951         to scalar_int_mode.
18952         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
18953         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
18954         (reg_num_sign_bit_copies_for_combine): Likewise.
18955         * rtlanal.c (nonzero_bits1): Likewise.
18956         (num_sign_bit_copies1): Likewise.
18957         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
18958         (reg_num_sign_bit_copies_general): Likewise.
18959         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
18960         (reg_nonzero_bits_general): Likewise.
18962 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18963             Alan Hayward  <alan.hayward@arm.com>
18964             David Sherwood  <david.sherwood@arm.com>
18966         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
18967         than in subroutines.  Return 1 for non-integer modes.
18968         (cached_num_sign_bit_copies): Change the type of the mode parameter
18969         to scalar_int_mode.
18970         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
18971         classes.  Handle CONST_INT_P first and then check whether X also
18972         has a scalar integer mode.  Check the same thing for inner registers
18973         of a SUBREG and for values that are being extended or truncated.
18975 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18976             Alan Hayward  <alan.hayward@arm.com>
18977             David Sherwood  <david.sherwood@arm.com>
18979         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
18980         in subroutines.  Return the mode mask for non-integer modes.
18981         (cached_nonzero_bits): Change the type of the mode parameter
18982         to scalar_int_mode.
18983         (nonzero_bits1): Likewise.  Remove early exit for other mode
18984         classes.  Handle CONST_INT_P first and then check whether X
18985         also has a scalar integer mode.
18987 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18988             Alan Hayward  <alan.hayward@arm.com>
18989             David Sherwood  <david.sherwood@arm.com>
18991         * expr.c (widest_int_mode_for_size): Make the comment match the code.
18992         Return a scalar_int_mode and assert that the size is greater than
18993         one byte.
18994         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
18995         (op_by_pieces_d::op_by_pieces_d): Likewise.
18996         (op_by_pieces_d::run): Likewise.
18997         (can_store_by_pieces): Likewise.
18999 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19000             Alan Hayward  <alan.hayward@arm.com>
19001             David Sherwood  <david.sherwood@arm.com>
19003         * combine.c (extract_left_shift): Add a mode argument and update
19004         recursive calls.
19005         (make_compound_operation_int): Change the type of the mode parameter
19006         to scalar_int_mode and update the call to extract_left_shift.
19008 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19009             Alan Hayward  <alan.hayward@arm.com>
19010             David Sherwood  <david.sherwood@arm.com>
19012         * combine.c (simplify_and_const_int): Change the type of the mode
19013         parameter to scalar_int_mode.
19014         (simplify_and_const_int_1): Likewise.  Update recursive call.
19016 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19017             Alan Hayward  <alan.hayward@arm.com>
19018             David Sherwood  <david.sherwood@arm.com>
19020         * combine.c (simplify_compare_const): Check that the mode is a
19021         scalar_int_mode (rather than VOIDmode) before testing its
19022         precision.
19023         (simplify_comparison): Move COMPARISON_P handling out of the
19024         loop and restrict the latter part of the loop to scalar_int_modes.
19025         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19026         and when considering SUBREG_REGs.  Use is_int_mode instead of
19027         checking GET_MODE_CLASS against MODE_INT.
19029 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19030             Alan Hayward  <alan.hayward@arm.com>
19031             David Sherwood  <david.sherwood@arm.com>
19033         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19034         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19035         shift_unit_mode and for modes involved in scalar shifts.
19037 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19038             Alan Hayward  <alan.hayward@arm.com>
19039             David Sherwood  <david.sherwood@arm.com>
19041         * combine.c (force_int_to_mode): New function, split out from...
19042         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19043         instead of GET_MODE (x).
19045 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19046             Alan Hayward  <alan.hayward@arm.com>
19047             David Sherwood  <david.sherwood@arm.com>
19049         * optabs-query.h (extraction_insn::struct_mode): Change type to
19050         opt_scalar_int_mode and update comment.
19051         (extraction_insn::field_mode): Change type to scalar_int_mode.
19052         (extraction_insn::pos_mode): Likewise.
19053         * combine.c (make_extraction): Update accordingly.
19054         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19055         (get_optab_extraction_insn): Likewise.
19056         * recog.c (simplify_while_replacing): Likewise.
19057         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19058         parameter to opt_scalar_int_mode.
19060 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19061             Alan Hayward  <alan.hayward@arm.com>
19062             David Sherwood  <david.sherwood@arm.com>
19064         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19065         to a scalar_int_mode instead of a machine_mode.
19066         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19067         (get_best_mode): Return a boolean and use a pointer argument to store
19068         the selected mode.  Replace the limit mode parameter with a bit limit.
19069         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19070         for the values returned by bit_field_mode_iterator::next_mode.
19071         (store_bit_field): Update call to get_best_mode.
19072         (store_fixed_bit_field): Likewise.
19073         (extract_fixed_bit_field): Likewise.
19074         * expr.c (optimize_bitfield_assignment_op): Likewise.
19075         * fold-const.c (optimize_bit_field_compare): Likewise.
19076         (fold_truth_andor_1): Likewise.
19077         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19078         Update for new type of m_mode.
19079         (get_best_mode): As above.
19081 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19082             Alan Hayward  <alan.hayward@arm.com>
19083             David Sherwood  <david.sherwood@arm.com>
19085         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19086         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19087         (store_bit_field): Check is_a <scalar_int_mode> before calling
19088         strict_volatile_bitfield_p.
19089         (extract_bit_field): Likewise.
19091 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19092             Alan Hayward  <alan.hayward@arm.com>
19093             David Sherwood  <david.sherwood@arm.com>
19095         * target.def (cstore_mode): Return a scalar_int_mode.
19096         * doc/tm.texi: Regenerate.
19097         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19098         * targhooks.h (default_cstore_mode): Likewise.
19099         * targhooks.c (default_cstore_mode): Likewise, using a forced
19100         conversion.
19101         * expmed.c (emit_cstore): Expect the target of the cstore to be
19102         a scalar_int_mode.
19104 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19105             Alan Hayward  <alan.hayward@arm.com>
19106             David Sherwood  <david.sherwood@arm.com>
19108         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19109         scalar_int_mode.
19110         (niter_desc): Likewise mode.
19111         (iv_analyze): Add a mode parameter.
19112         (biv_p): Likewise.
19113         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19114         and change its type to scalar_int_mode.
19115         * loop-iv.c: Update commentary at head of file.
19116         (iv_constant): Pass the mode paraeter before the rtx it describes
19117         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19118         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19119         (iv_extend): Likewise.
19120         (shorten_into_mode): Likewise.
19121         (iv_add): Use scalar_int_mode.
19122         (iv_mult): Likewise.
19123         (iv_shift): Likewise.
19124         (canonicalize_iv_subregs): Likewise.
19125         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19126         it describes and change its mode to scalar_int_mode.   Also change
19127         the type of the returned inner_mode to scalar_int_mode.
19128         (get_biv_step): Likewise, turning outer_mode from a pointer
19129         into a direct parameter.  Update call to get_biv_step_1.
19130         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19131         iv_constant and get_biv_step.
19132         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19133         and change its type to scalar_int_mode.  Don't initialise iv->mode
19134         to VOIDmode and remove later checks for its still being VOIDmode.
19135         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19136         is_a <scalar_int_mode> when changing the mode under consideration.
19137         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19138         Update call to iv_analyze_expr.
19139         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19140         inner register is not also a scalar_int_mode.  Update call to
19141         iv_analyze_biv.
19142         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19143         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19144         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19145         separate mode class checks.  Update calls to iv_analyze.  Remove
19146         fix-up of VOIDmodes after iv_analyze_biv.
19147         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19148         don't have a scalar_int_mode.  Update call to biv_p.
19150 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19151             Alan Hayward  <alan.hayward@arm.com>
19152             David Sherwood  <david.sherwood@arm.com>
19154         * cfgexpand.c (convert_debug_memory_address): Use
19155         as_a <scalar_int_mode>.
19156         * combine.c (expand_compound_operation): Likewise.
19157         (make_extraction): Likewise.
19158         (change_zero_ext): Likewise.
19159         (simplify_comparison): Likewise.
19160         * cse.c (cse_insn): Likewise.
19161         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19162         (mem_loc_descriptor): Likewise.
19163         (loc_descriptor): Likewise.
19164         * expmed.c (init_expmed_one_mode): Likewise.
19165         (synth_mult): Likewise.
19166         (emit_store_flag_1): Likewise.
19167         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19168         of a comparison with size.
19169         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19170         (reduce_to_bit_field_precision): Likewise.
19171         * function.c (expand_function_end): Likewise.
19172         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19173         * loop-doloop.c (doloop_modify): Likewise.
19174         * optabs.c (expand_binop): Likewise.
19175         (expand_unop): Likewise.
19176         (expand_copysign_absneg): Likewise.
19177         (prepare_cmp_insn): Likewise.
19178         (maybe_legitimize_operand): Likewise.
19179         * recog.c (const_scalar_int_operand): Likewise.
19180         * rtlanal.c (get_address_mode): Likewise.
19181         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19182         (simplify_cond_clz_ctz): Likewise.
19183         * tree-nested.c (get_nl_goto_field): Likewise.
19184         * tree.c (build_vector_type_for_mode): Likewise.
19185         * var-tracking.c (use_narrower_mode): Likewise.
19187 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19188             Alan Hayward  <alan.hayward@arm.com>
19189             David Sherwood  <david.sherwood@arm.com>
19191         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19192         * builtins.c (expand_builtin_signbit): Use it.
19193         * cfgexpand.c (expand_debug_expr): Likewise.
19194         * dojump.c (do_jump): Likewise.
19195         (do_compare_and_jump): Likewise.
19196         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19197         * expmed.c (make_tree): Likewise.
19198         * expr.c (expand_expr_real_2): Likewise.
19199         (expand_expr_real_1): Likewise.
19200         (try_casesi): Likewise.
19201         * fold-const-call.c (fold_const_call_ss): Likewise.
19202         * fold-const.c (unextend): Likewise.
19203         (extract_muldiv_1): Likewise.
19204         (fold_single_bit_test): Likewise.
19205         (native_encode_int): Likewise.
19206         (native_encode_string): Likewise.
19207         (native_interpret_int): Likewise.
19208         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19209         * internal-fn.c (expand_addsub_overflow): Likewise.
19210         (expand_neg_overflow): Likewise.
19211         (expand_mul_overflow): Likewise.
19212         (expand_arith_overflow): Likewise.
19213         * match.pd: Likewise.
19214         * stor-layout.c (layout_type): Likewise.
19215         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19216         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19217         * tree-ssanames.c (get_range_info): Likewise.
19218         * tree-switch-conversion.c (array_value_type) Likewise.
19219         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19220         (vect_recog_divmod_pattern): Likewise.
19221         (vect_recog_mixed_size_cond_pattern): Likewise.
19222         * tree-vrp.c (extract_range_basic): Likewise.
19223         (simplify_float_conversion_using_ranges): Likewise.
19224         * tree.c (int_fits_type_p): Likewise.
19225         * ubsan.c (instrument_bool_enum_load): Likewise.
19226         * varasm.c (mergeable_string_section): Likewise.
19227         (narrowing_initializer_constant_valid_p): Likewise.
19228         (output_constant): Likewise.
19230 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19231             Alan Hayward  <alan.hayward@arm.com>
19232             David Sherwood  <david.sherwood@arm.com>
19234         * machmode.h (NARROWEST_INT_MODE): New macro.
19235         * expr.c (alignment_for_piecewise_move): Use it instead of
19236         GET_CLASS_NARROWEST_MODE (MODE_INT).
19237         (push_block): Likewise.
19238         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19239         Likewise.
19240         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19242 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19243             Alan Hayward  <alan.hayward@arm.com>
19244             David Sherwood  <david.sherwood@arm.com>
19246         * postreload.c (move2add_valid_value_p): Change the type of the
19247         mode parameter to scalar_int_mode.
19248         (move2add_use_add2_insn): Add a mode parameter and use it instead
19249         of GET_MODE (reg).
19250         (move2add_use_add3_insn): Likewise.
19251         (reload_cse_move2add): Update accordingly.
19253 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19254             Alan Hayward  <alan.hayward@arm.com>
19255             David Sherwood  <david.sherwood@arm.com>
19257         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19258         double-word mode.
19259         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19260         * optabs.c (expand_unop): Likewise.
19262 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19263             Alan Hayward  <alan.hayward@arm.com>
19264             David Sherwood  <david.sherwood@arm.com>
19266         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19267         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19268         (popcount_loc_descriptor): Likewise.
19269         (bswap_loc_descriptor): Likewise.
19270         (rotate_loc_descriptor): Likewise.
19271         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19272         calling the functions above.
19274 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19275             Alan Hayward  <alan.hayward@arm.com>
19276             David Sherwood  <david.sherwood@arm.com>
19278         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19279         checks.
19280         (try_combine): Likewise.
19281         (simplify_if_then_else): Likewise.
19282         * cse.c (cse_insn): Likewise.
19283         * dwarf2out.c (mem_loc_descriptor): Likewise.
19284         * emit-rtl.c (gen_lowpart_common): Likewise.
19285         * simplify-rtx.c (simplify_truncation): Likewise.
19286         (simplify_binary_operation_1): Likewise.
19287         (simplify_const_relational_operation): Likewise.
19288         (simplify_ternary_operation): Likewise.
19289         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19291 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19292             Alan Hayward  <alan.hayward@arm.com>
19293             David Sherwood  <david.sherwood@arm.com>
19295         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19296         * reload.c (push_reload): Likewise.
19297         (find_reloads): Likewise.
19299 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19300             Alan Hayward  <alan.hayward@arm.com>
19301             David Sherwood  <david.sherwood@arm.com>
19303         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19304         (make_compound_operation_int): Likewise.
19305         (change_zero_ext): Likewise.
19306         * expr.c (convert_move): Likewise.
19307         (convert_modes): Likewise.
19308         * fwprop.c (forward_propagate_subreg): Likewise.
19309         * loop-iv.c (get_biv_step_1): Likewise.
19310         * optabs.c (widen_operand): Likewise.
19311         * postreload.c (move2add_valid_value_p): Likewise.
19312         * recog.c (simplify_while_replacing): Likewise.
19313         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19314         (simplify_binary_operation_1): Likewise.  Remove redundant
19315         mode equality check.
19317 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19318             Alan Hayward  <alan.hayward@arm.com>
19319             David Sherwood  <david.sherwood@arm.com>
19321         * combine.c (combine_simplify_rtx): Add checks for
19322         is_a <scalar_int_mode>.
19323         (simplify_if_then_else): Likewise.
19324         (make_field_assignment): Likewise.
19325         (simplify_comparison): Likewise.
19326         * ifcvt.c (noce_try_bitop): Likewise.
19327         * loop-invariant.c (canonicalize_address_mult): Likewise.
19328         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19330 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19331             Alan Hayward  <alan.hayward@arm.com>
19332             David Sherwood  <david.sherwood@arm.com>
19334         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19335         is_a <scalar_int_mode> instead of != BLKmode.
19337 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19338             Alan Hayward  <alan.hayward@arm.com>
19339             David Sherwood  <david.sherwood@arm.com>
19341         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19342         instead of != VOIDmode.
19343         * combine.c (if_then_else_cond): Likewise.
19344         (change_zero_ext): Likewise.
19345         * dwarf2out.c (mem_loc_descriptor): Likewise.
19346         (loc_descriptor): Likewise.
19347         * rtlanal.c (canonicalize_condition): Likewise.
19348         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19350 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19351             Alan Hayward  <alan.hayward@arm.com>
19352             David Sherwood  <david.sherwood@arm.com>
19354         * simplify-rtx.c (simplify_binary_operation_1): Use
19355         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19357 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19358             Alan Hayward  <alan.hayward@arm.com>
19359             David Sherwood  <david.sherwood@arm.com>
19361         * wide-int.h (int_traits<unsigned char>) New class.
19362         (int_traits<unsigned short>) Likewise.
19363         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19364         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19365         SCALAR_INT_MODE_P.
19366         * combine.c (set_nonzero_bits_and_sign_copies): Use
19367         is_a <scalar_int_mode>.
19368         (find_split_point): Likewise.
19369         (combine_simplify_rtx): Likewise.
19370         (simplify_logical): Likewise.
19371         (expand_compound_operation): Likewise.
19372         (expand_field_assignment): Likewise.
19373         (make_compound_operation): Likewise.
19374         (extended_count): Likewise.
19375         (change_zero_ext): Likewise.
19376         (simplify_comparison): Likewise.
19377         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19378         (ucompare_loc_descriptor): Likewise.
19379         (minmax_loc_descriptor): Likewise.
19380         (mem_loc_descriptor): Likewise.
19381         (loc_descriptor): Likewise.
19382         * expmed.c (init_expmed_one_mode): Likewise.
19383         * lra-constraints.c (lra_constraint_offset): Likewise.
19384         * optabs.c (prepare_libcall_arg): Likewise.
19385         * postreload.c (move2add_note_store): Likewise.
19386         * reload.c (operands_match_p): Likewise.
19387         * rtl.h (load_extend_op): Likewise.
19388         * rtlhooks.c (gen_lowpart_general): Likewise.
19389         * simplify-rtx.c (simplify_truncation): Likewise.
19390         (simplify_unary_operation_1): Likewise.
19391         (simplify_binary_operation_1): Likewise.
19392         (simplify_const_binary_operation): Likewise.
19393         (simplify_const_relational_operation): Likewise.
19394         (simplify_subreg): Likewise.
19395         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19396         * var-tracking.c (adjust_mems): Likewise.
19397         (prepare_call_arguments): Likewise.
19399 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19400             Alan Hayward  <alan.hayward@arm.com>
19401             David Sherwood  <david.sherwood@arm.com>
19403         * machmode.h (is_int_mode): New fuction.
19404         * combine.c (find_split_point): Use it.
19405         (combine_simplify_rtx): Likewise.
19406         (simplify_if_then_else): Likewise.
19407         (simplify_set): Likewise.
19408         (simplify_shift_const_1): Likewise.
19409         (simplify_comparison): Likewise.
19410         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19411         * cse.c (notreg_cost): Likewise.
19412         (cse_insn): Likewise.
19413         * cselib.c (cselib_lookup_1): Likewise.
19414         * dojump.c (do_jump_1): Likewise.
19415         (do_compare_rtx_and_jump): Likewise.
19416         * dse.c (get_call_args): Likewise.
19417         * dwarf2out.c (rtl_for_decl_init): Likewise.
19418         (native_encode_initializer): Likewise.
19419         * expmed.c (emit_store_flag_1): Likewise.
19420         (emit_store_flag): Likewise.
19421         * expr.c (convert_modes): Likewise.
19422         (store_field): Likewise.
19423         (expand_expr_real_1): Likewise.
19424         * fold-const.c (fold_read_from_constant_string): Likewise.
19425         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19426         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19427         * optabs.c (expand_binop): Likewise.
19428         (expand_unop): Likewise.
19429         (expand_abs_nojump): Likewise.
19430         (expand_one_cmpl_abs_nojump): Likewise.
19431         * simplify-rtx.c (mode_signbit_p): Likewise.
19432         (val_signbit_p): Likewise.
19433         (val_signbit_known_set_p): Likewise.
19434         (val_signbit_known_clear_p): Likewise.
19435         (simplify_relational_operation_1): Likewise.
19436         * tree.c (vector_type_mode): Likewise.
19438 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19439             Alan Hayward  <alan.hayward@arm.com>
19440             David Sherwood  <david.sherwood@arm.com>
19442         * machmode.h (smallest_mode_for_size): Fix formatting.
19443         (smallest_int_mode_for_size): New function.
19444         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19445         instead of smallest_mode_for_size.
19446         * combine.c (make_extraction): Likewise.
19447         * config/arc/arc.c (arc_expand_movmem): Likewise.
19448         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19449         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19450         * config/s390/s390.c (s390_expand_insv): Likewise.
19451         * config/sparc/sparc.c (assign_int_registers): Likewise.
19452         * config/spu/spu.c (spu_function_value): Likewise.
19453         (spu_function_arg): Likewise.
19454         * coverage.c (get_gcov_type): Likewise.
19455         (get_gcov_unsigned_t): Likewise.
19456         * dse.c (find_shift_sequence): Likewise.
19457         * expmed.c (store_bit_field_1): Likewise.
19458         * expr.c (convert_move): Likewise.
19459         (store_field): Likewise.
19460         * internal-fn.c (expand_arith_overflow): Likewise.
19461         * optabs-query.c (get_best_extraction_insn): Likewise.
19462         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19463         * stor-layout.c (layout_type): Likewise.
19464         (initialize_sizetypes): Likewise.
19465         * targhooks.c (default_get_mask_mode): Likewise.
19466         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19468 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19469             Alan Hayward  <alan.hayward@arm.com>
19470             David Sherwood  <david.sherwood@arm.com>
19472         * machmode.h (opt_mode::else_blk): New function.
19473         (int_mode_for_mode): Declare.
19474         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19475         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19476         return type.
19477         * cfgexpand.c (expand_debug_expr): Likewise.
19478         * combine.c (gen_lowpart_or_truncate): Likewise.
19479         (gen_lowpart_for_combine): Likewise.
19480         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19481         * config/avr/avr.c (avr_to_int_mode): Likewise.
19482         (avr_out_plus_1): Likewise.
19483         (avr_out_plus): Likewise.
19484         (avr_out_round): Likewise.
19485         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19486         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19487         (s390_expand_vcond): Likewise.
19488         * config/spu/spu.c (spu_split_immediate): Likewise.
19489         (spu_expand_mov): Likewise.
19490         * dse.c (get_stored_val): Likewise.
19491         * expmed.c (store_bit_field_1): Likewise.
19492         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19493         int_mode_for_size.
19494         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19495         (extract_low_bits): Likewise.
19496         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19497         handling rather than repeating the check.
19498         (emit_group_store): Likewise.
19499         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19500         * optabs.c (expand_absneg_bit): Likewise.
19501         (expand_copysign_absneg): Likewise.
19502         (expand_copysign_bit): Likewise.
19503         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19504         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19505         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19506         * var-tracking.c (prepare_call_arguments):  Likewise.
19507         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19508         int_mode_for_mode instead of mode_for_size.
19509         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19511 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19512             Alan Hayward  <alan.hayward@arm.com>
19513             David Sherwood  <david.sherwood@arm.com>
19515         * machmode.h (int_mode_for_size): New function.
19516         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19517         instead of mode_for_size.
19518         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19519         explicit.
19520         * combine.c (expand_field_assignment): Use int_mode_for_size
19521         instead of mode_for_size.
19522         (make_extraction): Likewise.
19523         (simplify_shift_const_1): Likewise.
19524         (simplify_comparison): Likewise.
19525         * dojump.c (do_jump): Likewise.
19526         * dwarf2out.c (mem_loc_descriptor): Likewise.
19527         * emit-rtl.c (init_derived_machine_modes): Likewise.
19528         * expmed.c (flip_storage_order): Likewise.
19529         (convert_extracted_bit_field): Likewise.
19530         * expr.c (copy_blkmode_from_reg): Likewise.
19531         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19532         * internal-fn.c (expand_mul_overflow): Likewise.
19533         * lower-subreg.c (simple_move): Likewise.
19534         * optabs-libfuncs.c (init_optabs): Likewise.
19535         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19536         * tree.c (vector_type_mode): Likewise.
19537         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19538         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19539         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19540         * tree-vect-stmts.c (vectorizable_load): Likewise.
19541         (vectorizable_store): Likewise.
19543 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19544             Alan Hayward  <alan.hayward@arm.com>
19545             David Sherwood  <david.sherwood@arm.com>
19547         * coretypes.h (pod_mode): New type.
19548         (scalar_int_mode_pod): New typedef.
19549         * machmode.h (pod_mode): New class.
19550         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19551         * genmodes.c (emit_mode_int_n): Update accordingly.
19552         * lower-subreg.h (target_lower_subreg): Change type to
19553         scalar_int_mode_pod.
19554         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19555         scalar_int_mode_pod.
19557 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19558             Alan Hayward  <alan.hayward@arm.com>
19559             David Sherwood  <david.sherwood@arm.com>
19561         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19562         machine_mode to scalar_int_mode.
19563         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19564         (rs6000_option_override_internal): Remove cast to int.
19565         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19566         machine_mode to scalar_int_mode.
19567         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19568         (rs6000_option_override_internal): Remove cast to int.
19569         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19570         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19571         to machine_mode.
19572         * config/s390/s390.c (s390_expand_builtin): Likewise.
19573         * coretypes.h (scalar_int_mode): New type.
19574         (opt_scalar_int_mode): New typedef.
19575         * machmode.h (scalar_int_mode): New class.
19576         (scalar_int_mode::includes_p): New function.
19577         (byte_mode): Change type to scalar_int_mode.
19578         (word_mode): Likewise.
19579         (ptr_mode): Likewise.
19580         * emit-rtl.c (byte_mode): Likewise.
19581         (word_mode): Likewise.
19582         (ptr_mode): Likewise.
19583         (init_derived_machine_modes): Update accordingly.
19584         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19585         and MODE_PARTIAL_INT.
19586         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19587         opt_scalar_int_mode.
19589 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19590             Alan Hayward  <alan.hayward@arm.com>
19591             David Sherwood  <david.sherwood@arm.com>
19593         * target.def (libgcc_floating_mode_supported_p): Take a
19594         scalar_float_mode.
19595         * doc/tm.texi: Regenerate.
19596         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19597         scalar_float_mode.
19598         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19599         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19600         Likewise.
19602 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19603             Alan Hayward  <alan.hayward@arm.com>
19604             David Sherwood  <david.sherwood@arm.com>
19606         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19607         * doc/tm.texi: Regenerate.
19608         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19609         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19610         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19611         * targhooks.h (default_floatn_mode): Likewise.
19612         * targhooks.c (default_floatn_mode): Likewise.
19613         * tree.c (build_common_tree_nodes): Update accordingly.
19615 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19616             Alan Hayward  <alan.hayward@arm.com>
19617             David Sherwood  <david.sherwood@arm.com>
19619         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19620         (mode_iterator::iterate_p): Likewise.
19621         (mode_iterator::get_wider): Likewise.
19622         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19624 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19625             Alan Hayward  <alan.hayward@arm.com>
19626             David Sherwood  <david.sherwood@arm.com>
19628         * coretypes.h (opt_scalar_float_mode): New typedef.
19629         * machmode.h (float_mode_for_size): New function.
19630         * emit-rtl.c (double_mode): Delete.
19631         (init_emit_once): Use float_mode_for_size.
19632         * stor-layout.c (layout_type): Likewise.
19633         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19635 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19636             Alan Hayward  <alan.hayward@arm.com>
19637             David Sherwood  <david.sherwood@arm.com>
19639         * output.h (assemble_real): Take a scalar_float_mode.
19640         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19641         * config/arm/arm.md (consttable_4): Likewise.
19642         (consttable_8): Likewise.
19643         (consttable_16): Likewise.
19644         * config/mips/mips.md (consttable_float): Likewise.
19645         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19646         * varasm.c (assemble_real): Take a scalar_float_mode.
19647         (output_constant_pool_2): Update accordingly.
19648         (output_constant): Likewise.
19650 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19651             Alan Hayward  <alan.hayward@arm.com>
19652             David Sherwood  <david.sherwood@arm.com>
19654         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19655         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19656         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19657         (native_encode_real): Likewise.
19658         (native_interpret_real): Likewise.
19659         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19660         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19662 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19663             Alan Hayward  <alan.hayward@arm.com>
19664             David Sherwood  <david.sherwood@arm.com>
19666         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19667         <scalar_float_mode>.  Simplify.
19668         (gen_extend_conv_libfunc): Likewise.
19670 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19671             Alan Hayward  <alan.hayward@arm.com>
19672             David Sherwood  <david.sherwood@arm.com>
19674         * coretypes.h (scalar_float_mode): New type.
19675         * machmode.h (mode_traits::from_int): Use machine_mode if
19676         USE_ENUM_MODES is defined.
19677         (is_a): New function.
19678         (as_a): Likewise.
19679         (dyn_cast): Likewise.
19680         (scalar_float_mode): New class.
19681         (scalar_float_mode::includes_p): New function.
19682         (is_float_mode): Likewise.
19683         * gdbhooks.py (MachineModePrinter): New class.
19684         (build_pretty_printer): Use it for scalar_float_mode.
19685         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19686         (format_helper::format_helper): Turn into a template.
19687         * genmodes.c (get_mode_class): New function.
19688         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19689         or machine_mode if none.
19690         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19691         as_a <scalar_float_mode>.
19692         * dwarf2out.c (mem_loc_descriptor): Likewise.
19693         (insert_float): Likewise.
19694         (add_const_value_attribute): Likewise.
19695         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19696         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19697         (expand_unop): Update accordingly.
19698         (expand_abs_nojump): Likewise.
19699         (expand_copysign_absneg): Take a scalar_float_mode.
19700         (expand_copysign_bit): Likewise.
19701         (expand_copysign): Update accordingly.
19703 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19704             Alan Hayward  <alan.hayward@arm.com>
19705             David Sherwood  <david.sherwood@arm.com>
19707         * coretypes.h (opt_mode): New class.
19708         * machmode.h (opt_mode): Likewise.
19709         (opt_mode::else_void): New function.
19710         (opt_mode::require): Likewise.
19711         (opt_mode::exists): Likewise.
19712         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19713         (GET_MODE_2XWIDER_MODE): Likewise.
19714         (mode_iterator::get_wider): Update accordingly.
19715         (mode_iterator::get_2xwider): Likewise.
19716         (mode_iterator::get_known_wider): Likewise, turning into a template.
19717         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19718         forcing a wider mode to exist.
19719         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19720         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19721         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19722         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19723         * lower-subreg.c (init_lower_subreg): Likewise.
19724         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19725         on the final iteration.
19726         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19727         a wider mode exists before asking for a move pattern.
19728         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19729         forcing a wider mode to exist.
19730         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19731         returning false if no such mode exists.
19732         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19733         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19734         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19735         Avoid checking for a MODE_INT if we already know the mode is not a
19736         SCALAR_INT_MODE_P.
19737         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19738         forcing a wider mode to exist.
19739         (expmed_mult_highpart_optab): Likewise.
19740         (expmed_mult_highpart): Likewise.
19741         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19742         using else_void.
19743         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19744         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19745         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19746         * internal-fn.c (expand_mul_overflow): Update use of
19747         GET_MODE_2XWIDER_MODE.
19748         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19749         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19750         GET_MODE_WIDER_MODE.
19751         (convert_plusminus_to_widen): Likewise.
19752         * tree-switch-conversion.c (array_value_type): Likewise.
19753         * var-tracking.c (emit_note_insn_var_location): Likewise.
19754         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19755         Return false inside rather than outside the loop if no wider mode
19756         exists
19757         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19758         and GET_MODE_2XWIDER_MODE
19759         (can_compare_p): Use else_void.
19760         * gdbhooks.py (OptMachineModePrinter): New class.
19761         (build_pretty_printer): Use it for opt_mode.
19763 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19764             Alan Hayward  <alan.hayward@arm.com>
19765             David Sherwood  <david.sherwood@arm.com>
19767         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19768         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19770 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19771             Alan Hayward  <alan.hayward@arm.com>
19772             David Sherwood  <david.sherwood@arm.com>
19774         * machmode.h (mode_traits): New structure.
19775         (get_narrowest_mode): New function.
19776         (mode_iterator::start): Likewise.
19777         (mode_iterator::iterate_p): Likewise.
19778         (mode_iterator::get_wider): Likewise.
19779         (mode_iterator::get_known_wider): Likewise.
19780         (mode_iterator::get_2xwider): Likewise.
19781         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19782         (FOR_EACH_MODE): Likewise.
19783         (FOR_EACH_MODE_FROM): Likewise.
19784         (FOR_EACH_MODE_UNTIL): Likewise.
19785         (FOR_EACH_WIDER_MODE): Likewise.
19786         (FOR_EACH_2XWIDER_MODE): Likewise.
19787         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19788         * combine.c (simplify_comparison): Likewise
19789         * config/i386/i386.c (type_natural_mode): Likewise.
19790         * cse.c (cse_insn): Likewise.
19791         * dse.c (find_shift_sequence): Likewise.
19792         * emit-rtl.c (init_derived_machine_modes): Likewise.
19793         (init_emit_once): Likewise.
19794         * explow.c (hard_function_value): Likewise.
19795         * expmed.c (extract_fixed_bit_field_1): Likewise.
19796         (extract_bit_field_1): Likewise.
19797         (expand_divmod): Likewise.
19798         (emit_store_flag_1): Likewise.
19799         * expr.c (init_expr_target): Likewise.
19800         (convert_move): Likewise.
19801         (alignment_for_piecewise_move): Likewise.
19802         (widest_int_mode_for_size): Likewise.
19803         (emit_block_move_via_movmem): Likewise.
19804         (copy_blkmode_to_reg): Likewise.
19805         (set_storage_via_setmem): Likewise.
19806         (compress_float_constant): Likewise.
19807         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19808         * optabs-query.c (get_best_extraction_insn): Likewise.
19809         * optabs.c (expand_binop): Likewise.
19810         (expand_twoval_unop): Likewise.
19811         (expand_twoval_binop): Likewise.
19812         (widen_leading): Likewise.
19813         (widen_bswap): Likewise.
19814         (expand_parity): Likewise.
19815         (expand_unop): Likewise.
19816         (prepare_cmp_insn): Likewise.
19817         (prepare_float_lib_cmp): Likewise.
19818         (expand_float): Likewise.
19819         (expand_fix): Likewise.
19820         (expand_sfix_optab): Likewise.
19821         * postreload.c (move2add_use_add2_insn): Likewise.
19822         * reg-stack.c (reg_to_stack): Likewise.
19823         * reginfo.c (choose_hard_reg_mode): Likewise.
19824         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19825         * stor-layout.c (mode_for_size): Likewise.
19826         (smallest_mode_for_size): Likewise.
19827         (mode_for_vector): Likewise.
19828         (finish_bitfield_representative): Likewise.
19829         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19830         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19831         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19832         * var-tracking.c (prepare_call_arguments): Likewise.
19834 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19835             Alan Hayward  <alan.hayward@arm.com>
19836             David Sherwood  <david.sherwood@arm.com>
19838         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19839         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19840         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19841         * machmode.h (mode_size): Move earlier in file.
19842         (mode_precision): Likewise.
19843         (mode_inner): Likewise.
19844         (mode_nunits): Likewise.
19845         (mode_unit_size): Likewise.
19846         (unit_unit_precision): Likewise.
19847         (mode_wider): Likewise.
19848         (mode_2xwider): Likewise.
19849         (machine_mode): New class.
19850         (mode_to_bytes): New function.
19851         (mode_to_bits): Likewise.
19852         (mode_to_precision): Likewise.
19853         (mode_to_inner): Likewise.
19854         (mode_to_unit_size): Likewise.
19855         (mode_to_unit_precision): Likewise.
19856         (mode_to_nunits): Likewise.
19857         (GET_MODE_SIZE): Use mode_to_bytes.
19858         (GET_MODE_BITSIZE): Use mode_to_bits.
19859         (GET_MODE_PRECISION): Use mode_to_precision.
19860         (GET_MODE_INNER): Use mode_to_inner.
19861         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19862         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
19863         (GET_MODE_NUNITS): Use mode_to_nunits.
19864         * system.h (ALWAYS_INLINE): New macro.
19865         * config/powerpcspe/powerpcspe-c.c
19866         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
19867         int for arg1_mode and arg2_mode.
19869 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19870             Alan Hayward  <alan.hayward@arm.com>
19871             David Sherwood  <david.sherwood@arm.com>
19873         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
19874         Prefix mode names with E_ in case statements.
19875         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19876         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
19877         (aarch64_split_simd_move): Likewise.
19878         (aarch64_gen_storewb_pair): Likewise.
19879         (aarch64_gen_loadwb_pair): Likewise.
19880         (aarch64_gen_store_pair): Likewise.
19881         (aarch64_gen_load_pair): Likewise.
19882         (aarch64_get_condition_code_1): Likewise.
19883         (aarch64_constant_pool_reload_icode): Likewise.
19884         (get_rsqrte_type): Likewise.
19885         (get_rsqrts_type): Likewise.
19886         (get_recpe_type): Likewise.
19887         (get_recps_type): Likewise.
19888         (aarch64_gimplify_va_arg_expr): Likewise.
19889         (aarch64_simd_container_mode): Likewise.
19890         (aarch64_emit_load_exclusive): Likewise.
19891         (aarch64_emit_store_exclusive): Likewise.
19892         (aarch64_expand_compare_and_swap): Likewise.
19893         (aarch64_gen_atomic_cas): Likewise.
19894         (aarch64_emit_bic): Likewise.
19895         (aarch64_emit_atomic_swap): Likewise.
19896         (aarch64_emit_atomic_load_op): Likewise.
19897         (aarch64_evpc_trn): Likewise.
19898         (aarch64_evpc_uzp): Likewise.
19899         (aarch64_evpc_zip): Likewise.
19900         (aarch64_evpc_ext): Likewise.
19901         (aarch64_evpc_rev): Likewise.
19902         (aarch64_evpc_dup): Likewise.
19903         (aarch64_gen_ccmp_first): Likewise.
19904         (aarch64_gen_ccmp_next): Likewise.
19905         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
19906         (alpha_emit_xfloating_libcall): Likewise.
19907         (emit_insxl): Likewise.
19908         (alpha_arg_type): Likewise.
19909         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
19910         (arc_preferred_simd_mode): Likewise.
19911         (arc_secondary_reload): Likewise.
19912         (get_arc_condition_code): Likewise.
19913         (arc_print_operand): Likewise.
19914         (arc_legitimate_constant_p): Likewise.
19915         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19916         * config/arc/arc.md (casesi_load): Likewise.
19917         (casesi_compact_jump): Likewise.
19918         * config/arc/predicates.md (proper_comparison_operator): Likewise.
19919         (cc_use_register): Likewise.
19920         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19921         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
19922         (arm_init_iwmmxt_builtins): Likewise.
19923         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
19924         (neon_expand_vector_init): Likewise.
19925         (arm_attr_length_move_neon): Likewise.
19926         (maybe_get_arm_condition_code): Likewise.
19927         (arm_emit_vector_const): Likewise.
19928         (arm_preferred_simd_mode): Likewise.
19929         (arm_output_iwmmxt_tinsr): Likewise.
19930         (thumb1_output_casesi): Likewise.
19931         (thumb2_output_casesi): Likewise.
19932         (arm_emit_load_exclusive): Likewise.
19933         (arm_emit_store_exclusive): Likewise.
19934         (arm_expand_compare_and_swap): Likewise.
19935         (arm_evpc_neon_vuzp): Likewise.
19936         (arm_evpc_neon_vzip): Likewise.
19937         (arm_evpc_neon_vrev): Likewise.
19938         (arm_evpc_neon_vtrn): Likewise.
19939         (arm_evpc_neon_vext): Likewise.
19940         (arm_validize_comparison): Likewise.
19941         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
19942         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
19943         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
19944         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
19945         (c6x_preferred_simd_mode): Likewise.
19946         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
19947         (epiphany_rtx_costs): Likewise.
19948         * config/epiphany/predicates.md (proper_comparison_operator):
19949         Likewise.
19950         * config/frv/frv.c (condexec_memory_operand): Likewise.
19951         (frv_emit_move): Likewise.
19952         (output_move_single): Likewise.
19953         (output_condmove_single): Likewise.
19954         (frv_hard_regno_mode_ok): Likewise.
19955         (frv_matching_accg_mode): Likewise.
19956         * config/h8300/h8300.c (split_adds_subs): Likewise.
19957         (h8300_rtx_costs): Likewise.
19958         (h8300_print_operand): Likewise.
19959         (compute_mov_length): Likewise.
19960         (output_logical_op): Likewise.
19961         (compute_logical_op_length): Likewise.
19962         (compute_logical_op_cc): Likewise.
19963         (h8300_shift_needs_scratch_p): Likewise.
19964         (output_a_shift): Likewise.
19965         (compute_a_shift_length): Likewise.
19966         (compute_a_shift_cc): Likewise.
19967         (expand_a_rotate): Likewise.
19968         (output_a_rotate): Likewise.
19969         * config/i386/i386.c (classify_argument): Likewise.
19970         (function_arg_advance_32): Likewise.
19971         (function_arg_32): Likewise.
19972         (function_arg_64): Likewise.
19973         (function_value_64): Likewise.
19974         (ix86_gimplify_va_arg): Likewise.
19975         (ix86_legitimate_constant_p): Likewise.
19976         (put_condition_code): Likewise.
19977         (split_double_mode): Likewise.
19978         (ix86_avx256_split_vector_move_misalign): Likewise.
19979         (ix86_expand_vector_logical_operator): Likewise.
19980         (ix86_split_idivmod): Likewise.
19981         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
19982         (ix86_build_const_vector): Likewise.
19983         (ix86_build_signbit_mask): Likewise.
19984         (ix86_match_ccmode): Likewise.
19985         (ix86_cc_modes_compatible): Likewise.
19986         (ix86_expand_branch): Likewise.
19987         (ix86_expand_sse_cmp): Likewise.
19988         (ix86_expand_sse_movcc): Likewise.
19989         (ix86_expand_int_sse_cmp): Likewise.
19990         (ix86_expand_vec_perm_vpermi2): Likewise.
19991         (ix86_expand_vec_perm): Likewise.
19992         (ix86_expand_sse_unpack): Likewise.
19993         (ix86_expand_int_addcc): Likewise.
19994         (ix86_split_to_parts): Likewise.
19995         (ix86_vectorize_builtin_gather): Likewise.
19996         (ix86_vectorize_builtin_scatter): Likewise.
19997         (avx_vpermilp_parallel): Likewise.
19998         (inline_memory_move_cost): Likewise.
19999         (ix86_tieable_integer_mode_p): Likewise.
20000         (x86_maybe_negate_const_int): Likewise.
20001         (ix86_expand_vector_init_duplicate): Likewise.
20002         (ix86_expand_vector_init_one_nonzero): Likewise.
20003         (ix86_expand_vector_init_one_var): Likewise.
20004         (ix86_expand_vector_init_concat): Likewise.
20005         (ix86_expand_vector_init_interleave): Likewise.
20006         (ix86_expand_vector_init_general): Likewise.
20007         (ix86_expand_vector_set): Likewise.
20008         (ix86_expand_vector_extract): Likewise.
20009         (emit_reduc_half): Likewise.
20010         (ix86_emit_i387_round): Likewise.
20011         (ix86_mangle_type): Likewise.
20012         (ix86_expand_round_sse4): Likewise.
20013         (expand_vec_perm_blend): Likewise.
20014         (canonicalize_vector_int_perm): Likewise.
20015         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20016         (expand_vec_perm_1): Likewise.
20017         (expand_vec_perm_interleave3): Likewise.
20018         (expand_vec_perm_even_odd_pack): Likewise.
20019         (expand_vec_perm_even_odd_1): Likewise.
20020         (expand_vec_perm_broadcast_1): Likewise.
20021         (ix86_vectorize_vec_perm_const_ok): Likewise.
20022         (ix86_expand_vecop_qihi): Likewise.
20023         (ix86_expand_mul_widen_hilo): Likewise.
20024         (ix86_expand_sse2_abs): Likewise.
20025         (ix86_expand_pextr): Likewise.
20026         (ix86_expand_pinsr): Likewise.
20027         (ix86_preferred_simd_mode): Likewise.
20028         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20029         * config/i386/sse.md (*andnot<mode>3): Likewise.
20030         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20031         (*<code><mode>3): Likewise.
20032         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20033         (ia64_expand_atomic_op): Likewise.
20034         (ia64_arg_type): Likewise.
20035         (ia64_mode_to_int): Likewise.
20036         (ia64_scalar_mode_supported_p): Likewise.
20037         (ia64_vector_mode_supported_p): Likewise.
20038         (expand_vec_perm_broadcast): Likewise.
20039         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20040         (iq2000_function_arg_advance): Likewise.
20041         (iq2000_function_arg): Likewise.
20042         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20043         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20044         (m68k_libcall_value): Likewise.
20045         (m68k_function_value): Likewise.
20046         (sched_attr_op_type): Likewise.
20047         * config/mcore/mcore.c (mcore_output_move): Likewise.
20048         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20049         Likewise.
20050         (microblaze_function_arg): Likewise.
20051         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20052         (mips_print_operand): Likewise.
20053         (mips_mode_ok_for_mov_fmt_p): Likewise.
20054         (mips_vector_mode_supported_p): Likewise.
20055         (mips_preferred_simd_mode): Likewise.
20056         (mips_expand_vpc_loongson_even_odd): Likewise.
20057         (mips_expand_vec_unpack): Likewise.
20058         (mips_expand_vi_broadcast): Likewise.
20059         (mips_expand_vector_init): Likewise.
20060         (mips_expand_vec_reduc): Likewise.
20061         (mips_expand_msa_cmp): Likewise.
20062         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20063         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20064         (cc_flags_for_mode): Likewise.
20065         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20066         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20067         (nds32_output_casesi_pc_relative): Likewise.
20068         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20069         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20070         (nvptx_gen_unpack): Likewise.
20071         (nvptx_gen_pack): Likewise.
20072         (nvptx_gen_shuffle): Likewise.
20073         (nvptx_gen_wcast): Likewise.
20074         (nvptx_preferred_simd_mode): Likewise.
20075         * config/pa/pa.c (pa_secondary_reload): Likewise.
20076         * config/pa/predicates.md (base14_operand): Likewise.
20077         * config/powerpcspe/powerpcspe-c.c
20078         (altivec_resolve_overloaded_builtin): Likewise.
20079         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20080         Likewise.
20081         (rs6000_preferred_simd_mode): Likewise.
20082         (output_vec_const_move): Likewise.
20083         (rs6000_expand_vector_extract): Likewise.
20084         (rs6000_split_vec_extract_var): Likewise.
20085         (reg_offset_addressing_ok_p): Likewise.
20086         (rs6000_legitimate_offset_address_p): Likewise.
20087         (rs6000_legitimize_address): Likewise.
20088         (rs6000_emit_set_const): Likewise.
20089         (rs6000_const_vec): Likewise.
20090         (rs6000_emit_move): Likewise.
20091         (spe_build_register_parallel): Likewise.
20092         (rs6000_darwin64_record_arg_recurse): Likewise.
20093         (swap_selector_for_mode): Likewise.
20094         (spe_init_builtins): Likewise.
20095         (paired_init_builtins): Likewise.
20096         (altivec_init_builtins): Likewise.
20097         (do_load_for_compare): Likewise.
20098         (rs6000_generate_compare): Likewise.
20099         (rs6000_expand_float128_convert): Likewise.
20100         (emit_load_locked): Likewise.
20101         (emit_store_conditional): Likewise.
20102         (rs6000_output_function_epilogue): Likewise.
20103         (rs6000_handle_altivec_attribute): Likewise.
20104         (rs6000_function_value): Likewise.
20105         (emit_fusion_gpr_load): Likewise.
20106         (emit_fusion_p9_load): Likewise.
20107         (emit_fusion_p9_store): Likewise.
20108         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20109         (fusion_gpr_mem_load): Likewise.
20110         (fusion_addis_mem_combo_load): Likewise.
20111         (fusion_addis_mem_combo_store): Likewise.
20112         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20113         (fusion_gpr_mem_load): Likewise.
20114         (fusion_addis_mem_combo_load): Likewise.
20115         (fusion_addis_mem_combo_store): Likewise.
20116         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20117         Likewise.
20118         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20119         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20120         (rs6000_preferred_simd_mode): Likewise.
20121         (output_vec_const_move): Likewise.
20122         (rs6000_expand_vector_extract): Likewise.
20123         (rs6000_split_vec_extract_var): Likewise.
20124         (reg_offset_addressing_ok_p): Likewise.
20125         (rs6000_legitimate_offset_address_p): Likewise.
20126         (rs6000_legitimize_address): Likewise.
20127         (rs6000_emit_set_const): Likewise.
20128         (rs6000_const_vec): Likewise.
20129         (rs6000_emit_move): Likewise.
20130         (rs6000_darwin64_record_arg_recurse): Likewise.
20131         (swap_selector_for_mode): Likewise.
20132         (paired_init_builtins): Likewise.
20133         (altivec_init_builtins): Likewise.
20134         (rs6000_expand_float128_convert): Likewise.
20135         (emit_load_locked): Likewise.
20136         (emit_store_conditional): Likewise.
20137         (rs6000_output_function_epilogue): Likewise.
20138         (rs6000_handle_altivec_attribute): Likewise.
20139         (rs6000_function_value): Likewise.
20140         (emit_fusion_gpr_load): Likewise.
20141         (emit_fusion_p9_load): Likewise.
20142         (emit_fusion_p9_store): Likewise.
20143         * config/rx/rx.c (rx_gen_move_template): Likewise.
20144         (flags_from_mode): Likewise.
20145         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20146         (s390_slb_comparison): Likewise.
20147         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20148         (s390_vector_mode_supported_p): Likewise.
20149         (s390_cc_modes_compatible): Likewise.
20150         (s390_match_ccmode_set): Likewise.
20151         (s390_canonicalize_comparison): Likewise.
20152         (s390_emit_compare_and_swap): Likewise.
20153         (s390_branch_condition_mask): Likewise.
20154         (s390_rtx_costs): Likewise.
20155         (s390_secondary_reload): Likewise.
20156         (__SECONDARY_RELOAD_CASE): Likewise.
20157         (s390_expand_cs): Likewise.
20158         (s390_preferred_simd_mode): Likewise.
20159         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20160         * config/sh/sh.c (sh_print_operand): Likewise.
20161         (dump_table): Likewise.
20162         (sh_secondary_reload): Likewise.
20163         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20164         * config/sh/sh.md (casesi_worker_1): Likewise.
20165         (casesi_worker_2): Likewise.
20166         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20167         (fcc_comparison_operator): Likewise.
20168         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20169         (emit_soft_tfmode_cvt): Likewise.
20170         (sparc_preferred_simd_mode): Likewise.
20171         (output_cbranch): Likewise.
20172         (sparc_print_operand): Likewise.
20173         (sparc_expand_vec_perm_bmask): Likewise.
20174         (vector_init_bshuffle): Likewise.
20175         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20176         (spu_vector_mode_supported_p): Likewise.
20177         (spu_expand_insv): Likewise.
20178         (spu_emit_branch_or_set): Likewise.
20179         (spu_handle_vector_attribute): Likewise.
20180         (spu_builtin_splats): Likewise.
20181         (spu_builtin_extract): Likewise.
20182         (spu_builtin_promote): Likewise.
20183         (spu_expand_sign_extend): Likewise.
20184         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20185         (tilegx_simd_int): Likewise.
20186         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20187         (tilepro_simd_int): Likewise.
20188         * config/v850/v850.c (const_double_split): Likewise.
20189         (v850_print_operand): Likewise.
20190         (ep_memory_offset): Likewise.
20191         * config/vax/vax.c (vax_rtx_costs): Likewise.
20192         (vax_output_int_move): Likewise.
20193         (vax_output_int_add): Likewise.
20194         (vax_output_int_subtract): Likewise.
20195         * config/visium/predicates.md (visium_branch_operator): Likewise.
20196         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20197         (visium_print_operand_address): Likewise.
20198         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20199         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20200         (xtensa_expand_conditional_branch): Likewise.
20201         (xtensa_copy_incoming_a7): Likewise.
20202         (xtensa_output_literal): Likewise.
20203         * dfp.c (decimal_real_maxval): Likewise.
20204         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20206 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20207             Alan Hayward  <alan.hayward@arm.com>
20208             David Sherwood  <david.sherwood@arm.com>
20210         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20211         (mode_nunits_inline): Likewise.
20212         (mode_inner_inline): Likewise.
20213         (mode_unit_size_inline): Likewise.
20214         (mode_unit_precision_inline): Likewise.
20215         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20216         unprefixed name.
20217         (emit_mode_wider): Add an E_ prefix to mode names.
20218         (emit_mode_complex): Likewise.
20219         (emit_mode_inner): Likewise.
20220         (emit_mode_adjustments): Likewise.
20221         (emit_mode_int_n): Likewise.
20222         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20223         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20224         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20225         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20226         (CRC32_BUILTIN, ENTRY): Likewise.
20227         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20228         (aarch64_pop_regs): Likewise.
20229         (aarch64_process_components): Likewise.
20230         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20231         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20232         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20233         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20234         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20235         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20236         Likewise.
20237         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20238         (pcmpistr): Likewise.
20239         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20240         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20241         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20242         Likewise.
20243         * config/rl78/rl78.c (mduc_regs): Likewise.
20244         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20245         (htm_expand_builtin): Likewise.
20246         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20247         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20248         * config/xtensa/xtensa.c (print_operand): Likewise.
20249         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20250         (NUM_MODE_VECTOR_INT): Likewise.
20251         * genoutput.c (null_operand): Likewise.
20252         (output_operand_data): Likewise.
20253         * genrecog.c (print_parameter_value): Likewise.
20254         * lra.c (debug_operand_data): Likewise.
20256 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20258         * dwarf2out.c (scompare_loc_descriptor_wide)
20259         (scompare_loc_descriptor_narrow): New functions, split out from...
20260         (scompare_loc_descriptor): ...here.
20261         * expmed.c (emit_store_flag_int): New function, split out from...
20262         (emit_store_flag): ...here.
20264 2017-08-30  Richard Biener  <rguenther@suse.de>
20266         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20267         (dwarf2out_early_finish): Move setting of AT_pubnames from
20268         early debug output to early finish.
20270 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20272         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20273         and -mdata-region to the assembler.
20275 2017-08-30  Richard Biener  <rguenther@suse.de>
20277         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20278         attributes.
20279         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20280         (dwarf2out_early_global_decl): Only generate a DIE for the
20281         abstract origin if it doesn't already exist or is a declaration DIE.
20282         (resolve_addr): Do not add the linkage name twice when
20283         generating a stub DIE for the DW_TAG_GNU_call_site target.
20285 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20287         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20288         Use machine_mode rather than int for arg1_mode.
20290 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20292         PR target/82015
20293         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20294         that the second argument of the built-in functions to unpack
20295         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20296         switch statement instead a lot of if statements.
20297         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20298         Allow 64-bit values to be in Altivec registers as well as
20299         traditional floating point registers.
20300         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20302 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20304         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20305         MAX_REGS_PER_ADDRESS == 1.
20307 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20309         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20310         * config/i386/i386.c (ix86_option_override_internal): Simplify
20311         setting of opts->x_flag_entry.
20313 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20314             Jakub Jelinek  <jakub@redhat.com>
20315             Richard Biener  <rguenther@suse.de>
20317         PR tree-optimization/81503
20318         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20319         folded constant fits in the target type; reorder tests for clarity.
20321 2017-08-29  Martin Liska  <mliska@suse.cz>
20323         * passes.def: Include pass_lower_switch.
20324         * stmt.c (dump_case_nodes): Remove and move to
20325         tree-switch-conversion.
20326         (case_values_threshold): Likewise.
20327         (expand_switch_as_decision_tree_p): Likewise.
20328         (emit_case_decision_tree): Likewise.
20329         (expand_case): Likewise.
20330         (balance_case_nodes): Likewise.
20331         (node_has_low_bound): Likewise.
20332         (node_has_high_bound): Likewise.
20333         (node_is_bounded): Likewise.
20334         (emit_case_nodes): Likewise.
20335         (struct simple_case_node): New struct.
20336         (add_case_node): Remove.
20337         (emit_case_dispatch_table): Use vector instead of case_list.
20338         (reset_out_edges_aux): Remove.
20339         (compute_cases_per_edge): Likewise.
20340         (expand_case): Build list of simple_case_node.
20341         (expand_sjlj_dispatch_table): Use it.
20342         * tree-switch-conversion.c (struct case_node): Moved from
20343         stmt.c and adjusted.
20344         (emit_case_nodes): Likewise.
20345         (node_has_low_bound): Likewise.
20346         (node_has_high_bound): Likewise.
20347         (node_is_bounded): Likewise.
20348         (case_values_threshold): Likewise.
20349         (reset_out_edges_aux): Likewise.
20350         (compute_cases_per_edge): Likewise.
20351         (add_case_node): Likewise.
20352         (dump_case_nodes): Likewise.
20353         (balance_case_nodes): Likewise.
20354         (expand_switch_as_decision_tree_p): Likewise.
20355         (emit_jump): Likewise.
20356         (emit_case_decision_tree): Likewise.
20357         (try_switch_expansion): Likewise.
20358         (do_jump_if_equal): Likewise.
20359         (emit_cmp_and_jump_insns): Likewise.
20360         (fix_phi_operands_for_edge): New function.
20361         (record_phi_operand_mapping): Likewise.
20362         (class pass_lower_switch): New pass.
20363         (pass_lower_switch::execute): New function.
20364         (make_pass_lower_switch): Likewise.
20365         (conditional_probability):
20366         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20367         * tree-pass.h: Add make_pass_lower_switch.
20369 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20371         PR target/80993
20372         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20373         handlers as used.
20375 2017-08-29  Richard Biener  <rguenther@suse.de>
20377         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20378         we do not add a DW_AT_inline attribute twice.
20379         (gen_subprogram_die): Remove code setting DW_AT_inline on
20380         DECL_ABSTRACT_P nodes.
20382 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20384         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20385         calls to internal functions.
20386         (gimplify_modify_expr): Likewise.
20387         * tree-call-cdce.c (use_internal_fn): Likewise.
20388         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20389         (convert_to_divmod): Set the nothrow flag.
20390         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20391         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20392         (vectorizable_call): Likewise.
20393         (vectorizable_store): Likewise.
20394         (vectorizable_load): Likewise.
20395         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20396         (vect_recog_mask_conversion_pattern): Likewise.
20398 2017-08-29  Martin Liska  <mliska@suse.cz>
20400         PR other/39851
20401         * gcc.c (driver_handle_option): Add new argument.
20402         * opts-common.c (handle_option): Pass
20403         target_option_override_hook.
20404         * opts-global.c (lang_handle_option): Add new option.
20405         (set_default_handlers):  Add new argument.
20406         (decode_options): Likewise.
20407         * opts.c (target_handle_option): Likewise.
20408         (common_handle_option): Call target_option_override_hook.
20409         * opts.h (struct cl_option_handler_func): Add hook for
20410         target option override.
20411         (struct cl_option_handlers): Likewise.
20412         (set_default_handlers): Add new argument.
20413         (decode_options): Likewise.
20414         (common_handle_option): Likewise.
20415         (target_handle_option): Likewise.
20416         * toplev.c (toplev::main): Pass targetm.target_option.override
20417         hook.
20419 2017-08-29  Richard Biener  <rguenther@suse.de>
20420         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20422         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20423         life to the active subtree.
20425 2017-08-28  Jeff Law  <law@redhat.com>
20427         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20428         derive_equivalences.
20429         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20430         Code moved into....
20431         (edge_info::derive_equivalences): New private member function
20433         * tree-ssa-dom.c (class edge_info): Changed from a struct
20434         to a class.  Add ctor/dtor, methods and data members.
20435         (edge_info::edge_info): Renamed from allocate_edge_info.
20436         Initialize additional members.
20437         (edge_info::~edge_info): New.
20438         (free_dom_edge_info): Delete the edge info.
20439         (record_edge_info): Use new class & associated member functions.
20440         Tighten forms for testing for edge equivalences.
20441         (record_temporary_equivalences): Iterate over the simple
20442         equivalences rather than assuming there's only one per edge.
20443         (cprop_into_successor_phis): Iterate over the simple
20444         equivalences rather than assuming there's only one per edge.
20445         (optimize_stmt): Use operand_equal_p rather than pointer
20446         equality for mini-DSE code.
20448 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20450         * gcc.c (execute): Fold SIGPIPE handling into switch
20451         statement.  Adjust internal error message.
20453 2017-08-28  Richard Biener  <rguenther@suse.de>
20455         PR debug/81993
20456         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20457         Do nothing for removed DIEs.
20459 2017-08-28  Richard Biener  <rguenther@suse.de>
20461         PR tree-optimization/81977
20462         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20463         memcpy.
20465 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20467         PR target/80640
20468         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20469         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20470         using targetm.gen_mem_thread_fence.
20472 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20474         PR target/81995
20475         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20476         predicate to register_operand.  Reorder operands.
20477         (*btr<mode>): Ditto.
20478         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20479         (*btr<mode>_mask): Ditto.
20481 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20483         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20484         * config/rs6000/xmmintrin.h: New file.
20485         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20487 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20489         PR target/81504
20490         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20491         parameter and_insn and return it.
20492         (recombine_lvx_pattern): Insert a copy to ensure availability of
20493         the base register of the copied masking operation at the point of
20494         the instruction replacement.
20495         (recombine_stvx_pattern): Likewise.
20497 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20499         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20500         undocumented switches.
20501         (-mpower9-dform-vector): Likewise.
20502         (-mpower9-dform): Likewise.
20503         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20504         comments to delete references to -mpower9-dform* switches.
20505         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20506         Delete reference to -mpower9-dform* switches, test for
20507         -mpower9-vector instead.
20508         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20509         (OTHER_P9_VECTOR_MASKS): Likewise.
20510         (POWERPC_MASKS): Likewise.
20511         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20512         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20513         that checked for -mpower9-dform* consistancy with other options.
20514         Add test for -mpower9-misc to enable other power9 switches.
20515         (rs6000_init_hard_regno_mode_ok): Likewise.
20516         (rs6000_option_override_internal): Likewise.
20517         (rs6000_emit_prologue): Likewise.
20518         (rs6000_emit_epilogue): Likewise.
20519         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20520         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20521         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20522         -mpower9-vector.
20523         (emit_fusion_p9_store): Likewise.
20524         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20525         resetting these macros if the assembler does not support ISA 3.0
20526         instructions.
20527         (TARGET_P9_DFORM_VECTOR): Likewise.
20528         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20529         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20531 2017-08-25  Alan Modra  <amodra@gmail.com>
20533         PR rtl-optimization/81747
20534         * cse.c (cse_extended_basic_block): Don't attempt to record
20535         equivalences for degenerate conditional branches that branch
20536         to their fall-through.
20538 2017-08-24  Martin Sebor  <msebor@redhat.com>
20540         PR middle-end/81908
20541         * gimple-fold.c (size_must_be_zero_p): New function.
20542         (gimple_fold_builtin_memory_op): Call it.
20544 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20546         * config/rs6000/mm_malloc.h: New file.
20548 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20550         PR tree-optimization/81913
20551         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20552         analysis when either IVs in condition can wrap.
20554 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20556         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20557         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20559 2017-08-24  Richard Biener  <rguenther@suse.de>
20561         PR target/81921
20562         * targhooks.c (default_target_can_inline_p): Properly
20563         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20564         is present and always compare.
20565         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20566         infer -mfpmath=sse from TARGET_SSE_P.
20567         (ix86_can_inline_p): Properly use target_option_default_node when
20568         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20570 2017-08-24  Richard Biener  <rguenther@suse.de>
20572         PR debug/81936
20573         * dwarf2out.c (output_die): Handle flag_generate_offload like
20574         flag_generate_lto.
20575         (output_comp_unit): Likewise.
20576         (gen_array_type_die): Likewise.
20577         (dwarf2out_early_finish): Likewise.
20578         (note_variable_value_in_expr): Likewise.
20579         (dwarf2out_finish): Likewise.  Adjust assert.
20580         * cgraphunit.c (symbol_table::compile): Move setting of
20581         flag_generate_offload earlier ...
20582         (symbol_table::finalize_compilation_unit): ... here, before
20583         early debug finalization.
20585 2017-08-24  Richard Biener  <rguenther@suse.de>
20587         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20588         and ipa-fnsummary.h.
20589         (ix86_can_inline_p): When ix86_fpmath flags do not match
20590         check whether the callee uses FP math at all.
20592 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20594         PR middle-end/81931
20595         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20596         instead of TYPE_PRECISION.
20598 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20599             Alan Hayward  <alan.hayward@arm.com>
20600             David Sherwood  <david.sherwood@arm.com>
20602         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20604 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20606         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20607         Remove field.
20608         (ix86_frame::stack_realign_allocate): New field.
20609         (struct machine_frame_state): Modify comments.
20610         (machine_frame_state::sp_realigned_fp_end): New field.
20611         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20612         layout calculation.
20613         (sp_valid_at): Add assertion to assure no attempt to access invalid
20614         offset of a realigned stack.
20615         (fp_valid_at): Likewise.
20616         (choose_baseaddr): Modify comments.
20617         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20618         ix86_expand_prologue.
20619         (ix86_expand_prologue): Modify stack realignment and allocation.
20620         (ix86_expand_epilogue): Modify comments.
20621         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20622         avx2_runtime, avx512f, and avx512f_runtime.
20624 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20626         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20627         (mstackrealign): Do not init to -1.
20628         * config/i386/i386.c (ix86_option_override_internal):
20629         Check opts_set, not opts when setting default value of
20630         opts->x_ix86_force_align_arg_pointer.
20632 2017-08-23  Richard Biener  <rguenther@suse.de>
20634         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20635         lang_hooks.decl_printable_name.
20636         * print-rtl-function.c (print_rtx_function): Likewise.
20637         * tree-pretty-print.c (dump_function_header): Likewise.
20639 2017-08-23  Richard Biener  <rguenther@suse.de>
20641         PR lto/81940
20642         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20643         -g0 at compile-time.
20645 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20647         PR middle-end/19706
20648         * doc/sourcebuild.texi (Other hardware attributes):
20649         Document xorsign.
20651 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20653         PR middle-end/19706
20654         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20655         Fix single-use check.
20657 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20659         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20661 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20663         * doc/install.texi: Modify to add more details on running selected
20664         tests.
20666 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20668         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20669         is combined with -mabi=ms.
20670         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20671         ms_abi.
20673 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20675         PR tree-optimization/81488
20676         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20677         and cached_basis fields.
20678         (MAX_SPREAD): New constant.
20679         (alloc_cand_and_find_basis): Initialize new fields.
20680         (clear_visited): New function.
20681         (create_phi_basis_1): Rename from create_phi_basis, set visited
20682         and cached_basis fields.
20683         (create_phi_basis): New wrapper function.
20684         (phi_add_costs_1): Rename from phi_add_costs, add spread
20685         parameter, set visited field, short-circuit when limits reached.
20686         (phi_add_costs): New wrapper function.
20687         (record_phi_increments_1): Rename from record_phi_increments, set
20688         visited field.
20689         (record_phi_increments): New wrapper function.
20690         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20691         (phi_incr_cost): New wrapper function.
20692         (all_phi_incrs_profitable_1): Rename from
20693         all_phi_incrs_profitable, set visited field.
20694         (all_phi_incrs_profitable): New wrapper function.
20696 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20697             Alan Hayward  <alan.hayward@arm.com>
20698             David Sherwood  <david.sherwood@arm.com>
20700         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20701         that takes the outer and inner modes.
20702         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20703         comparison as the canonical test for a paradoxical subreg.
20704         * combine.c (simplify_set): Use paradoxical_subreg_p.
20705         (make_extraction): Likewise.
20706         (force_to_mode): Likewise.
20707         (rtx_equal_for_field_assignment_p): Likewise.
20708         (gen_lowpart_for_combine): Likewise.
20709         (simplify_comparison): Likewise.
20710         * cse.c (equiv_constant): Likewise.
20711         * expmed.c (store_bit_field_1): Likewise.
20712         * final.c (alter_subreg): Likewise.
20713         * fwprop.c (propagate_rtx): Likewise.
20714         (forward_propagate_subreg): Likewise.
20715         * ira-conflicts.c (ira_build_conflicts): Likewise.
20716         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20717         * lra-constraints.c (curr_insn_transform): Likewise.
20718         (split_reg): Likewise.
20719         * lra-eliminations.c (move_plus_up): Likewise.
20720         (lra_eliminate_regs_1): Likewise.
20721         * recog.c (general_operand): Likewise.
20722         * ree.c (combine_reaching_defs): Likewise.
20723         * reload.c (push_reload): Likewise.
20724         (find_reloads): Likewise.
20725         * reload1.c (elimination_effects): Likewise.
20726         (compute_reload_subreg_offset): Likewise.
20727         (choose_reload_regs): Likewise.
20728         * rtlanal.c (subreg_lsb_1): Likewise.
20729         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20730         (simplify_subreg): Likewise.
20731         * var-tracking.c (track_loc_p): Likewise.
20732         * emit-rtl.c (byte_lowpart_offset): Likewise.
20733         (paradoxical_subreg_p): Delete out-of-line definition.
20735 2017-08-22  Jeff Law  <law@redhat.com>
20737         PR tree-optimization/81741
20738         PR tree-optimization/71947
20739         * tree-ssa-dom.c: Include tree-inline.h.
20740         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20741         equivalences if one is more expensive to compute than the other.
20742         * tree-ssa-scopedtables.h (class const_or_copies): Make
20743         record_const_or_copy_raw method private.
20744         (class avail_exprs_stack): New method simplify_binary_operation.
20745         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20746         avail_exprs_stack::simplify_binary_operation as needed.
20747         (avail_exprs_stack::simplify_binary_operation): New function.
20749 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20751         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20752         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20753         (DOT_SYMBOLS): Likewise.
20754         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20755         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20756         (RS6000_ABI_NAME): Likewise.
20757         (TARGET_CMODEL): Likewise.
20758         (TOC_SECTION_ASM_OP): Likewise.
20759         (SET_CMODEL): New macro.
20760         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20762 2017-08-22  Richard Biener  <rguenther@suse.de>
20764         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20765         to allow for free-lang-data replacements similar to verify_type_variant.
20767 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20768             Alan Hayward  <alan.hayward@arm.com>
20769             David Sherwood  <david.sherwood@arm.com>
20771         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20772         VOIDmode for the LABEL_REF.
20774 2017-08-22  Richard Biener  <rguenther@suse.de>
20776         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20778 2017-08-22  Richard Biener  <rguenther@suse.de>
20780         * common.opt (feliminate-dwarf2-dups): Ignore.
20781         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20782         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20783         same_die_p_wrap, compute_section_prefix,
20784         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20785         (comdat_symbol_id, comdat_symbol_number): Likewise.
20786         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20787         Likewise.
20788         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20789         (output_die): Mark unreachable path unreachable.
20790         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20791         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20792         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20793         (dwarf2out_early_finish): Likewise.
20795 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20797         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20799 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20801         PR target/81910
20802         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20803         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20804         (avr_attribute_table) <io, io_low, address>: Initialize
20805         .decl_required with true.
20807 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20809         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20810         undocumented debugging options.
20811         (-mvsx-scalar-double): Likewise.
20812         (-mallow-df-permute): Likewise.
20813         (-mvectorize-builtins): Likewise.
20814         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20815         (rs6000_builtin_vectorized_function): Likewise.
20816         (rs6000_builtin_md_vectorized_function): Likewise.
20817         (rs6000_opt_vars): Likewise.
20819 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20821         PR target/46091
20822         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20823         (*btrq_imm): Rename from *btrq.
20824         (*btcq_imm): Rename from *btcq.
20825         (btsc): New code attribute.
20826         (*<btsc><mode>): New insn pattern.
20827         (*btr<mode>): Ditto.
20828         (*<btsc><mode>_mask): New insn_and_split pattern.
20829         (*btr<mode>_mask): Ditto.
20831 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20832             Alan Hayward  <alan.hayward@arm.com>
20833             David Sherwood  <david.sherwood@arm.com>
20835         * function.c (pad_below): Simplify padding calculation.
20837 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20838             Alan Hayward  <alan.hayward@arm.com>
20839             David Sherwood  <david.sherwood@arm.com>
20841         * target.def (function_prologue): Remove frame size argument.
20842         (function_epilogue): Likewise.
20843         * doc/tm.texi: Regenerate.
20844         * final.c (final_start_function): Update call to function_prologue.
20845         (final_end_function): Update call to function_epilogue.
20846         (default_function_pro_epilogue): Remove frame size argument.
20847         * output.h (default_function_pro_epilogue): Likewise.
20848         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20849         (arm_output_function_prologue): Likewise.
20850         * config/frv/frv.c (frv_function_prologue): Likewise.
20851         (frv_function_epilogue): Likewise.
20852         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20853         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20854         (ia64_output_function_epilogue): Likewise.
20855         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20856         (m32r_output_function_epilogue): Likewise.
20857         * config/microblaze/microblaze.c (microblaze_function_prologue)
20858         (microblaze_function_epilogue): Likewise.
20859         * config/mips/mips.c (mips_output_function_prologue): Likewise.
20860         (mips_output_function_epilogue): Likewise.
20861         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20862         (mmix_target_asm_function_epilogue): Likewise.
20863         * config/msp430/msp430.c (msp430_start_function): Likewise.
20864         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
20865         (nds32_asm_function_epilogue): Likewise.
20866         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
20867         * config/pa/pa.c (pa_output_function_prologue): Likewise.
20868         (pa_output_function_epilogue): Likewise.
20869         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
20870         (rs6000_output_function_epilogue): Likewise.
20871         * config/rl78/rl78.c (rl78_start_function): Likewise.
20872         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
20873         (rs6000_output_function_epilogue): Likewise.
20874         * config/rx/rx.c (rx_output_function_prologue): Likewise.
20875         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
20876         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
20877         (sparc_asm_function_epilogue): Likewise.
20879 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20881         * tree.h (type_has_mode_precision_p): New function.
20882         * convert.c (convert_to_integer_1): Use it.
20883         * expr.c (expand_expr_real_2): Likewise.
20884         (expand_expr_real_1): Likewise.
20885         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
20886         * match.pd: Likewise.
20887         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
20888         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
20889         * tree-tailcall.c (process_assignment): Likewise.
20890         * tree-vect-loop.c (vectorizable_reduction): Likewise.
20891         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
20892         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
20893         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20894         (vectorizable_assignment): Likewise.
20895         (vectorizable_shift): Likewise.
20896         (vectorizable_operation): Likewise.
20897         * tree-vrp.c (register_edge_assert_for_2): Likewise.
20899 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
20901         * match.pd: Add pow (C, x) simplification.
20903 2017-08-21  Richard Biener  <rguenther@suse.de>
20905         PR tree-optimization/81900
20906         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
20907         for blocks with abnormal predecessors.
20908         (compute_antic): Do not set visited flag prematurely.
20910 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
20912         PR target/79883
20913         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
20915 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20917         * stor-layout.h (vector_type_mode): Move to...
20918         * tree.h (vector_type_mode): ...here.
20919         * stor-layout.c (vector_type_mode): Move to...
20920         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
20922 2017-08-21  Richard Biener  <rguenther@suse.de>
20924         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
20925         register_external_die hooks.
20926         (debug_false_tree_charstarstar_uhwistar): Declare.
20927         (debug_nothing_tree_charstar_uhwi): Likewise.
20928         * debug.c (do_nothing_debug_hooks): Adjust.
20929         (debug_false_tree_charstarstar_uhwistar): New do nothing.
20930         (debug_nothing_tree_charstar_uhwi): Likewise.
20931         * dbxout.c (dbx_debug_hooks): Adjust.
20932         (xcoff_debug_hooks): Likewise.
20933         * sdbout.c (sdb_debug_hooks): Likewise.
20934         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
20935         * dwarf2out.c (macinfo_label_base): New global.
20936         (dwarf2out_register_external_die): New function for the
20937         register_external_die hook.
20938         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
20939         (dwarf2_debug_hooks): Use them.
20940         (dwarf2_lineno_debug_hooks): Adjust.
20941         (struct die_struct): Add with_offset flag.
20942         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
20943         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
20944         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
20945         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
20946         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
20947         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
20948         defining section names for the early LTO debug variants.
20949         (reset_indirect_string): New helper.
20950         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
20951         (print_dw_val): Add support for offsetted symbol references.
20952         (get_ultimate_context): Split out from is_cxx.
20953         (is_cxx): Use get_ultimate_context.
20954         (is_fortran): Add decl overload.
20955         (compute_comp_unit_symbol): Split out worker from
20956         compute_section_prefix.
20957         (compute_section_prefix): Call compute_comp_unit_symbol and
20958         set comdat_type_p here.
20959         (output_die): Skip DIE symbol output for the LTO added one.
20960         Handle DIE symbol references with offset.
20961         (output_comp_unit): Guard section name mangling properly.
20962         For LTO debug sections emit a symbol at the section beginning
20963         which we use to refer to its DIEs.
20964         (add_abstract_origin_attribute): For DIEs registered via
20965         dwarf2out_register_external_die directly refer to the early
20966         DIE rather than indirectly through the shadow one we created.
20967         Remove obsolete call to dwarf2out_abstract_function for
20968         non-function/block origins.
20969         (gen_array_type_die): When generating early LTO debug do
20970         not emit DW_AT_string_length.
20971         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
20972         late when in LTO.  As suggested place a gcc_unreachable for
20973         the DECL_ABSTRACT_P case.
20974         (gen_subprogram_die): Avoid another specification DIE
20975         for early built declarations/definitions for the late LTO case.
20976         (gen_variable_die): Add type references for late duplicated VLA dies
20977         when in late LTO.
20978         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
20979         we have the abstract instance already.
20980         (process_scope_var): Adjust decl DIE contexts in LTO which
20981         first puts them in limbo.
20982         (gen_decl_die): Do not generate type DIEs late apart from
20983         types for VLAs or for decls we do not yet have a DIE.  Do not
20984         call dwarf2out_abstract_function late.
20985         (dwarf2out_early_global_decl): Make sure to create DIEs
20986         for abstract instances of a decl first.
20987         (dwarf2out_late_global_decl): Adjust comment.
20988         (output_macinfo_op): With multiple macro sections use
20989         macinfo_label_base to distinguish labels.
20990         (output_macinfo): Likewise.  Update macinfo_label_base.
20991         Pass in the line info label.
20992         (note_variable_value_in_expr): When generating LTO resolve
20993         all variable values here by generating DIEs as needed.
20994         (init_sections_and_labels): Add early LTO debug flag parameter
20995         and generate different sections and names if set.  Add generation
20996         counter for the labels so we can have multiple of them.
20997         (reset_dies): Helper to allow DIEs to be output multiple times.
20998         (dwarf2out_finish): When outputting DIEs to the fat part of an
20999         LTO object first reset DIEs.
21000         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21001         (modified_type_die): Check for decl_ultimate_origin being self
21002         before recursing.
21003         (gen_type_die_with_usage): Likewise.
21004         (gen_typedef_die): Allow decl_ultimate_origin being self.
21005         (set_decl_abstract_flags): Remove.
21006         (set_block_abstract_flags): Likewise.
21007         (dwarf2out_abstract_function): Treat the early generated DIEs
21008         as the abstract copy and only add DW_AT_inline and
21009         DW_AT_artificial here and call set_decl_origin_self.
21010         If the DIE has an abstract origin don't do anything.
21011         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21012         if we have none yet (Go fails to build one, PR78628).
21013         (variably_modified_type_p): Prevent endless recursion for Ada
21014         cyclic pointer types.
21015         * lto-streamer-in.c: Include debug.h.
21016         (dref_queue): New global.
21017         (lto_read_tree_1): Stream in DIE references.
21018         (lto_input_tree): Register DIE references.
21019         (input_function): Stream DECL_DEBUG_ARGS.
21020         * lto-streamer-out.c: Include debug.h.
21021         (lto_write_tree_1): Output DIE references.
21022         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21023         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21024         (output_function): Stream DECL_DEBUG_ARGS.
21025         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21026         Stream DECL_ABSTRACT_ORIGIN.
21027         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21028         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21029         DECL_CONTEXT for file-scope decls.
21030         * lto-streamer.h (struct dref_entry): Declare.
21031         (dref_queue): Likewise.
21032         * cfgexpand.c (pass_expand::execute): Do not call the
21033         outlining_inline_function hook here.
21034         * lto-wrapper.c (debug_obj): New global.
21035         (tool_cleanup): Unlink it if required.
21036         (debug_objcopy): New function.
21037         (run_gcc): Handle early debug sections in the IL files by
21038         extracting them to separate files, partially linkin them and
21039         feeding the result back as result to the linker.
21040         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21041         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21042         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21043         sections into a separate segment.
21044         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21045         segments.
21046         (darwin_asm_dwarf_section): Likewise.
21047         (darwin_asm_output_dwarf_offset): Likewise.
21048         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21050 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21051             Alan Hayward  <alan.hayward@arm.com>
21052             David Sherwood  <david.sherwood@arm.com>
21054         * read-md.h (md_reader::record_potential_iterator_use): Replace
21055         pointer argument with an rtx and an index.
21056         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21057         (apply_mode_iterator): Likewise.
21058         (apply_code_iterator): Likewise.
21059         (apply_int_iterator): Likewise.
21060         (apply_subst_iterator): Likewise.
21061         (record_iterator_use): Likewise.
21062         (record_attribute_use): Likewise.
21063         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21064         to record_iterator_use and apply_iterator.
21065         (iterator_use): Replace ptr with x and index.
21066         (attribute_use): Likewise.
21067         (apply_attribute_uses): Update calls to apply_iterator.
21068         (apply_iterators): Likewise.  Update initialization of iterator_use.
21069         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21070         and record_potential_iterator_use.
21071         (rtx_reader::read_rtx_operand): Likewise.
21073 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21074             Alan Hayward  <alan.hayward@arm.com>
21075             David Sherwood  <david.sherwood@arm.com>
21077         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21078         CONST_WIDE_INT.
21080 2017-08-21  Richard Biener  <rguenther@suse.de>
21082         PR middle-end/81884
21083         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21084         at struct end conservatively when comparing common bases.
21086 2017-08-21  Richard Biener  <rguenther@suse.de>
21088         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21089         (mem_ref_in_stmt): Remove.
21090         (determine_max_movement): Use ref index to get at the reference.
21091         (invariantness_dom_walker::before_dom_children): Deal with
21092         lim data already initialized.
21093         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21095 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21097         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21098         (TARGET_ISA_ROUND): Ditto.
21099         (TARGET_ROUND): Ditto.
21100         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21101         * config/i386/i386.md: Ditto.
21102         * config/i386/sse.md: Ditto.
21103         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21104         with OPTION_MASK_ISA_SSE4_1.
21106 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21108         PR target/81894
21109         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21110         __builtin_ia32_lzcnt_u16.
21112 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21114         PR target/80210
21115         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21116         (rs6000_set_current_function): Rewrite function to use it.
21118 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21120         PR c/53037
21121         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21122         and TYPE_WARN_IF_NOT_ALIGN.
21123         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21124         (handle_warn_if_not_align): New.
21125         (place_union_field): Call handle_warn_if_not_align.
21126         (place_field): Call handle_warn_if_not_align.
21127         Copy TYPE_WARN_IF_NOT_ALIGN.
21128         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21129         (layout_type): Likewise.
21130         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21131         spare to 18.
21132         (tree_decl_common): Add warn_if_not_align.
21133         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21134         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21135         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21136         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21137         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21138         * doc/extend.texi: Document warn_if_not_aligned attribute.
21139         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21141 2017-08-17  Martin Liska  <mliska@suse.cz>
21143         PR bootstrap/81864
21144         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21145         (get_data_dependence): Use it as pointer type.
21146         (distribute_loop): Likewise.
21148 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21150         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21151         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21152         (p8_vmrgow_<mode>_direct): New define_insn.
21153         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21154         handle endianness for vmrgew and vmrgow permute patterns.
21156 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21158         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21159         * config/rs6000/rs6000-cpus.def: Remove comment.
21160         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21161         (POWERPC_MASKS): Likewise.
21162         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21163         use of TARGET_VSX_TIMODE.
21164         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21165         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21166         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21167         (rs6000_option_override_internal): Remove dead code.
21168         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21169         (rs6000_legitimize_reload_address): Likewise.
21170         (rs6000_legitimate_address_p): Likewise.
21171         (rs6000_opt_masks): Delete "vsx-timode".
21172         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21173         from function comment.
21174         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21175         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21176         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21177         condition.
21178         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21179         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21180         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21181         (VSX_TI): Likewise.
21182         (VSX_M): Likewise.
21183         (define_peephole2): Likewise.
21185 2017-08-17  Martin Sebor  <msebor@redhat.com>
21187         PR c/81859
21188         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21189         past the end of an array.
21190         (test_pp_format): Add test cases.
21192 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21194         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21195         missing ECF_NOTHROW flags.
21197 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21199         PR target/72804
21200         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21201         operands residing in integer registers.
21202         (*vsx_le_perm_load_<mode>): Likewise.
21203         (*vsx_le_perm_store_<mode>): Likewise.
21204         (define_peephole2): Add peepholes to optimize the above.
21206 2017-08-17  Marek Polacek  <polacek@redhat.com>
21208         PR middle-end/81814
21209         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21210         to mimic what shorten_compare did.  Change the return type to bool.
21211         (fold_cond_expr_with_comparison): Update call to
21212         operand_equal_for_comparison_p.
21213         (fold_ternary_loc): Likewise.
21215 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21217         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21218         register.
21219         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21221 2017-08-17  Richard Biener  <rguenther@suse.de>
21223         * tree-ssa-structalias.c (solve_graph): When propagating
21224         to successors update the graphs succ edges and avoid duplicate work.
21226 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21228         PR target/81861
21229         * config/i386/i386.c (ix86_option_override_internal): Save target
21230         specific options after ix86_stack_protector_guard_reg was changed.
21232 2017-08-17  Richard Biener  <rguenther@suse.de>
21234         PR tree-optimization/81827
21235         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21236         (new_var_info): Initialize it conservatively.
21237         (get_call_vi): Mark register vars.
21238         (new_scalar_tmp_constraint_exp): Likewise.
21239         (handle_rhs_call): Likewise.
21240         (handle_const_call): Likewise.
21241         (create_function_info_for): Likewise.
21242         (solve_constraints): Sort varinfos to separate register from
21243         non-register vars to pack points-to solution bitmaps during
21244         iteration.
21246 2017-08-17  Marek Polacek  <polacek@redhat.com>
21248         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21250 2017-08-17  Richard Biener  <rguenther@suse.de>
21252         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21253         to true when overflow is undefined and we saturated the result.
21255 2017-08-17  Alan Modra  <amodra@gmail.com>
21257         PR target/80938
21258         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21259         Don't use store multiple if only one reg needs saving.
21260         (interesting_frame_related_regno): New function.
21261         (rs6000_frame_related): Don't emit frame info for regs that
21262         don't need saving.
21263         (rs6000_emit_epilogue): Likewise.
21265 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21267         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21268         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21269         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21270         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21271         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21272         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21273         (verify_type): Adjust for TYPE_BINFO move.
21274         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21275         process TYPE_BINFO directly.
21276         (hash_tree): Likewise.
21277         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21278         Likewise.
21279         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21280         Likewise.
21282 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21284         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21286 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21288         PR target/46091
21289         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21290         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21291         Add ix86_binary_operator_ok to insn constraint.
21292         (*iordi_1_bts): Ditto.
21293         (*xordi_1_btc): Ditto.
21294         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21295         Update corresponding peephole2 pattern.
21296         (*btrq): Ditto.
21297         (*btcq): Ditto.
21299 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21301         PR tree-optimization/81832
21302         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21303         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21305 2017-08-16  Marek Polacek  <polacek@redhat.com>
21307         PR middle/81695
21308         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21309         regarding size_zero_node.
21311 2017-08-16  Martin Liska  <mliska@suse.cz>
21313         PR target/81753
21314         * config.gcc: Respect previously set extra_objs in case
21315         of darwin target.
21317 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21319         PR tree-optimization/81835
21320         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21321         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21322         not depend on the phi.
21324 2017-08-16  Alan Modra  <amodra@gmail.com>
21326         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21327         dead code.
21329 2017-08-16  Alan Modra  <amodra@gmail.com>
21331         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21332         (save_reg_p): ..into this.  Update all callers.
21333         (first_reg_to_save): Simplify.
21335 2017-08-16  Alan Modra  <amodra@gmail.com>
21337         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21338         fixed regs.
21340 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21342         PR target/78460
21343         PR target/67712
21344         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21345         constant count if that count is less than 32.
21347 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21349         * gcc.c (execute): Emit friendlier message if inferior is killed
21350         by an external cause.
21352 2017-08-15  Richard Biener  <rguenther@suse.de>
21354         PR tree-optimization/81790
21355         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21356         CONSTRUCTORs from simplifying and VN.
21358 2017-08-14  Martin Sebor  <msebor@redhat.com>
21360         * builtin-attrs.def: Add comments.
21362 2017-08-14  Martin Sebor  <msebor@redhat.com>
21364         PR c/81117
21365         * doc/extend.texi (attribute nonstring): Document new attribute.
21367 2017-08-14  Martin Sebor  <msebor@redhat.com>
21369         PR c/81117
21370         * tree-diagnostic.c (default_tree_printer): Handle %G.
21371         * gimple-pretty-print.h (percent_G_format): Declare new function.
21372         * gimple-pretty-print.c (percent_G_format): Define.
21373         * tree-pretty-print.c (percent_K_format): Add argument.
21375 2017-08-14  Martin Sebor  <msebor@redhat.com>
21377         PR translation/79998
21378         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21379         Remove a stray space.
21381 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21383         PR target/46091
21384         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21385         (*iordi_1_bts): Ditto.
21386         (*xordi_1_btc): Ditto.
21388 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21390         PR target/79845
21391         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21392         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21393         Likewise.
21394         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21395         quoted strings, and make more translator-friendly.
21396         (darwin_rs6000_override_options): Likewise.
21397         (rs6000_option_override_internal): Likewise.
21398         (rs6000_return_in_memory): Fix overlong line.
21399         (init_cmulative_args): Use quoted strings, and make more
21400         translator-friendly.
21401         (rs6000_pass_by_reference): Fix overlong line.
21402         (def_builtin): Use quoted strings.
21403         (altivec_expand_predicate_builtin): Use quoted strings, and make
21404         more translator-friendly.
21405         (htm_expand_builtin): Use quoted strings.
21406         (cpu_expand_builtin): Use quoted strings, and make more
21407         translator-friendly.
21408         (altivec_expand_builtin): Likewise.
21409         (paired_expand_predicate_builtin): Likewise.
21410         (rs6000_invalid_builtin): Likewise.
21411         (builtin_function_type): Use quoted strings.
21412         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21413         more translator-friendly.
21414         (rs6000_trampoline_init): Likewise.
21415         (rs6000_handle_altivec_attribute): Likewise.
21416         (rs6000_inner_target_options): Use quoted strings.
21417         (rs6000_disable_incompatible_switches): Likewise.
21418         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21419         strings, and make more translator-friendly.
21420         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21422 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21424         PR tree-optimization/81799
21425         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21426         cond_expr to simple gimple operand.
21428 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21430         PR middle-end/46932
21431         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21433 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21435         PR target/81754
21436         PR target/81268
21437         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21438         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21439         TARGET_GASISR_PROLOGUES.
21440         * config/avr/avr.c (avr_option_override): Same.
21441         (avr_pass_pre_proep::execute): Same.
21443 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21445         PR target/81820
21446         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21447         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21449 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21451         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21452         define_insn_and_split.  Split to a memory load from 0 in
21453         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21454         using PTR mode iterator.
21455         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21456         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21457         (*add_tp_<mode>): Redefine as define_insn_and_split.
21458         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21459         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21460         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21461         Split to an add with a  memory load from 0 in
21462         DEFAULT_TLS_SEG_REG address space.
21464 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21466         * config/aarch64/aarch64-option-extensions.def (rdma):
21467         Fix feature string to what Linux prints out in /proc/cpuinfo.
21469 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21471         PR ada/79542
21472         * dwarf2out.c (modified_type_die): For C typedef types that have
21473         an ultimate origin, process the ultimate origin instead of the
21474         input type.
21475         (gen_typedef_die): Assert that input DECLs have no ultimate
21476         origin.
21477         (gen_type_die_with_usage): For typedef variants that have an
21478         ultimate origin, just call gen_decl_die on the original DECL.
21479         (process_scope_var): Avoid creating DIEs for local typedefs and
21480         concrete static variables.
21482 2017-08-12  Alan Modra  <amodra@gmail.com>
21484         PR target/81170
21485         PR target/81295
21486         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21487         match gnu-user.h startfile.
21488         (ENDFILE_LINUX_SPEC): Similarly.
21490 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21492         PR lto/81430
21493         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21494         Remove function.
21495         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21497 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21498         * config/aarch64/aarch64.md (mov<mode>): Change.
21499         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21500         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21501         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21503 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21505         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21506         for storage order barriers.
21508 2017-08-11  Martin Liska  <mliska@suse.cz>
21510         PR tree-opt/79987
21511         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21512         variables of void type.
21514 2017-08-11  Martin Liska  <mliska@suse.cz>
21516         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21517         TARGET_SUPPORTS_ALIASES.
21518         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21519         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21520         (optimize_weakref): Likewise.
21521         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21522         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21523         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21525 2017-08-11  Martin Liska  <mliska@suse.cz>
21527         PR ipa/81213
21528         * config/i386/i386.c (make_resolver_func): Do complete
21529         refactoring of the function.
21531 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21533         PR target/81708
21534         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21535         * config/i386/i386.c (ix86_stack_protect_guard): Use
21536         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21537         * doc/invoke.texi (x86 Options): Document
21538         -mstack-protector-guard-symbol= option.
21540 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21542         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21543         * config/i386/i386.c (ix86_split_stack_guard): New function.
21544         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21545         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21546         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21547         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21548         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21549         (split_stack_space_check): Call ix86_split_stack_guard.
21551 2017-08-10  Martin Sebor  <msebor@redhat.com>
21553         * print-tree.c (print_node): Print location using the established
21554         format %s:%i%i.
21555         Replace spaces with colons.
21556         (debug_raw, debug): Ditto.
21558 2017-08-10  Martin Sebor  <msebor@redhat.com>
21560         PR c++/81586
21561         * pretty-print.c (pp_format): Correct the handling of %s precision.
21563 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21565         PR target/81736
21566         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21567         to ...
21568         (ix86_finalize_stack_frame_flags): This.  Also clear
21569         frame_pointer_needed if -fno-omit-frame-pointer is used without
21570         stack access.
21571         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21572         with ix86_finalize_stack_frame_flags.
21573         (ix86_expand_epilogue): Likewise.
21574         (ix86_expand_split_stack_prologue): Likewise.
21575         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21577 2017-08-10  Martin Liska  <mliska@suse.cz>
21579         PR c++/81355
21580         * c-attribs.c (handle_target_attribute):
21581         Report warning for an empty string argument of target attribute.
21583 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21585         PR c/81687
21586         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21587         LABEL_DECLs.
21588         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21589         or DECL_NONLOCAL labels.
21590         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21591         or DECL_NONLOCAL labels here.
21593 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21595         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21596         to indicate when early gimple folding has been disabled.
21597         (rs6000_gimple_fold_builtin): Add debug content.
21598         (rs6000_invalid_builtin): Fix whitespace.
21599         (rs6000_expand_builtin): Fix whitespace.
21600         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21602 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21604         PR target/80938
21605         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21606         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21608 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21610         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21611         (qdf24xx): Likewise.
21612         * config/aarch64/aarch64.md: Include falkor.md.
21613         * config/aarch64/falkor.md: New.
21615 2017-08-09  Marek Polacek  <polacek@redhat.com>
21617         PR c/81233
21618         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21619         * diagnostic.c (emit_diagnostic): Add a comment.
21620         (emit_diagnostic_valist): New function.
21622 2017-08-09  Marek Polacek  <polacek@redhat.com>
21624         PR c/81417
21625         * input.c (make_location): New overload.
21626         * input.h (make_location): Declare.
21628 2017-08-08  Alan Modra  <amodra@gmail.com>
21629             H.J. Lu  <hongjiu.lu@intel.com>
21631         PR driver/81523
21632         * gcc.c (NO_PIE_SPEC): Delete.
21633         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21634         exclusion..
21635         (LINK_PIE_SPEC): ..to here.
21636         (LINK_COMMAND_SPEC): Support -no-pie.
21637         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21638         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21639         format.
21640         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21641         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21642         (ENDFILE_CRTEND_SPEC): Similarly.
21644 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21646         PR target/81708
21647         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21648         (mstack-protector-guard-offset=): Ditto.
21649         * config/i386/i386.c (ix86_option_override): Handle
21650         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21651         options.
21652         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21653         ix86_stack_protect_guard_offset variables.
21654         (TARGET_STACK_PROTECT_GUARD): Always define.
21655         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21656         and -mstack-protector-guard-offset= options.
21658 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21660         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21661         boundary case for the last candidate.
21663 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21665         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21666         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21668 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21670         PR middle-end/19706
21671         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21672         * config/aarch64/aarch64-builtins.c
21673         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21674         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21675         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21677 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21678             Andrew Pinski <pinskia@gmail.com>
21680         PR middle-end/19706
21681         * internal-fn.def (XORSIGN): New.
21682         * optabs.def (xorsign_optab): New.
21683         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21684         (convert_expand_mult_copysign): New.
21685         (pass_optimize_widening_mul::execute): Call
21686         convert_expand_mult_copysign.
21688 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21690         PR tree-optimization/81354
21691         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21692         Insert on edges rather than explicitly creating landing pads.
21693         (analyze_candidates_and_replace): Commit edge inserts.
21695 2017-08-08  Richard Biener  <rguenther@suse.de>
21697         PR middle-end/81719
21698         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21699         (expand_simple_operations): Also look through ADDR_EXPRs with
21700         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21702 2017-08-08  Richard Biener  <rguenther@suse.de>
21704         PR tree-optimization/81723
21705         * tree-vect-slp.c (struct bst_traits): New hash traits.
21706         (bst_fail): New global.
21707         (vect_build_slp_tree_2): New worker, split out from ...
21708         (vect_build_slp_tree): ... this now wrapping it with using
21709         bst_fail set to cache SLP tree build fails.  Properly handle
21710         max_tree_size.
21711         (vect_analyze_slp_instance): Allocate and free bst_fail.
21713 2017-08-08  Martin Liska  <mliska@suse.cz>
21715         PR tree-opt/81696
21716         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21717         LABEL_DECLs that can be from a different function.
21719 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21721         PR tree-optimization/81744
21722         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21723         loop's number of iterations.
21725 2017-08-08  Martin Liska  <mliska@suse.cz>
21727         * asan.c: Include header files.
21728         * attribs.c (build_decl_attribute_variant): New function moved
21729         from tree.[ch].
21730         (build_type_attribute_qual_variant): Likewise.
21731         (cmp_attrib_identifiers): Likewise.
21732         (simple_cst_list_equal): Likewise.
21733         (omp_declare_simd_clauses_equal): Likewise.
21734         (attribute_value_equal): Likewise.
21735         (comp_type_attributes): Likewise.
21736         (build_type_attribute_variant): Likewise.
21737         (lookup_ident_attribute): Likewise.
21738         (remove_attribute): Likewise.
21739         (merge_attributes): Likewise.
21740         (merge_type_attributes): Likewise.
21741         (merge_decl_attributes): Likewise.
21742         (merge_dllimport_decl_attributes): Likewise.
21743         (handle_dll_attribute): Likewise.
21744         (attribute_list_equal): Likewise.
21745         (attribute_list_contained): Likewise.
21746         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21747         (lookup_attribute_by_prefix): Likewise.
21748         * bb-reorder.c: Include header files.
21749         * builtins.c: Likewise.
21750         * calls.c: Likewise.
21751         * cfgexpand.c: Likewise.
21752         * cgraph.c: Likewise.
21753         * cgraphunit.c: Likewise.
21754         * convert.c: Likewise.
21755         * dwarf2out.c: Likewise.
21756         * final.c: Likewise.
21757         * fold-const.c: Likewise.
21758         * function.c: Likewise.
21759         * gimple-expr.c: Likewise.
21760         * gimple-fold.c: Likewise.
21761         * gimple-pretty-print.c: Likewise.
21762         * gimple.c: Likewise.
21763         * gimplify.c: Likewise.
21764         * hsa-common.c: Likewise.
21765         * hsa-gen.c: Likewise.
21766         * internal-fn.c: Likewise.
21767         * ipa-chkp.c: Likewise.
21768         * ipa-cp.c: Likewise.
21769         * ipa-devirt.c: Likewise.
21770         * ipa-fnsummary.c: Likewise.
21771         * ipa-inline.c: Likewise.
21772         * ipa-visibility.c: Likewise.
21773         * ipa.c: Likewise.
21774         * lto-cgraph.c: Likewise.
21775         * omp-expand.c: Likewise.
21776         * omp-general.c: Likewise.
21777         * omp-low.c: Likewise.
21778         * omp-offload.c: Likewise.
21779         * omp-simd-clone.c: Likewise.
21780         * opts-global.c: Likewise.
21781         * passes.c: Likewise.
21782         * predict.c: Likewise.
21783         * sancov.c: Likewise.
21784         * sanopt.c: Likewise.
21785         * symtab.c: Likewise.
21786         * toplev.c: Likewise.
21787         * trans-mem.c: Likewise.
21788         * tree-chkp.c: Likewise.
21789         * tree-eh.c: Likewise.
21790         * tree-into-ssa.c: Likewise.
21791         * tree-object-size.c: Likewise.
21792         * tree-parloops.c: Likewise.
21793         * tree-profile.c: Likewise.
21794         * tree-ssa-ccp.c: Likewise.
21795         * tree-ssa-live.c: Likewise.
21796         * tree-ssa-loop.c: Likewise.
21797         * tree-ssa-sccvn.c: Likewise.
21798         * tree-ssa-structalias.c: Likewise.
21799         * tree-ssa.c: Likewise.
21800         * tree-streamer-in.c: Likewise.
21801         * tree-vectorizer.c: Likewise.
21802         * tree-vrp.c: Likewise.
21803         * tsan.c: Likewise.
21804         * ubsan.c: Likewise.
21805         * varasm.c: Likewise.
21806         * varpool.c: Likewise.
21807         * tree.c: Remove functions moved to attribs.[ch].
21808         * tree.h: Likewise.
21809         * config/aarch64/aarch64.c: Add attrs.h header file.
21810         * config/alpha/alpha.c: Likewise.
21811         * config/arc/arc.c: Likewise.
21812         * config/arm/arm.c: Likewise.
21813         * config/avr/avr.c: Likewise.
21814         * config/bfin/bfin.c: Likewise.
21815         * config/c6x/c6x.c: Likewise.
21816         * config/cr16/cr16.c: Likewise.
21817         * config/cris/cris.c: Likewise.
21818         * config/darwin.c: Likewise.
21819         * config/epiphany/epiphany.c: Likewise.
21820         * config/fr30/fr30.c: Likewise.
21821         * config/frv/frv.c: Likewise.
21822         * config/ft32/ft32.c: Likewise.
21823         * config/h8300/h8300.c: Likewise.
21824         * config/i386/winnt.c: Likewise.
21825         * config/ia64/ia64.c: Likewise.
21826         * config/iq2000/iq2000.c: Likewise.
21827         * config/lm32/lm32.c: Likewise.
21828         * config/m32c/m32c.c: Likewise.
21829         * config/m32r/m32r.c: Likewise.
21830         * config/m68k/m68k.c: Likewise.
21831         * config/mcore/mcore.c: Likewise.
21832         * config/microblaze/microblaze.c: Likewise.
21833         * config/mips/mips.c: Likewise.
21834         * config/mmix/mmix.c: Likewise.
21835         * config/mn10300/mn10300.c: Likewise.
21836         * config/moxie/moxie.c: Likewise.
21837         * config/msp430/msp430.c: Likewise.
21838         * config/nds32/nds32-isr.c: Likewise.
21839         * config/nds32/nds32.c: Likewise.
21840         * config/nios2/nios2.c: Likewise.
21841         * config/nvptx/nvptx.c: Likewise.
21842         * config/pa/pa.c: Likewise.
21843         * config/pdp11/pdp11.c: Likewise.
21844         * config/powerpcspe/powerpcspe.c: Likewise.
21845         * config/riscv/riscv.c: Likewise.
21846         * config/rl78/rl78.c: Likewise.
21847         * config/rx/rx.c: Likewise.
21848         * config/s390/s390.c: Likewise.
21849         * config/sh/sh.c: Likewise.
21850         * config/sol2.c: Likewise.
21851         * config/sparc/sparc.c: Likewise.
21852         * config/spu/spu.c: Likewise.
21853         * config/stormy16/stormy16.c: Likewise.
21854         * config/tilegx/tilegx.c: Likewise.
21855         * config/tilepro/tilepro.c: Likewise.
21856         * config/v850/v850.c: Likewise.
21857         * config/vax/vax.c: Likewise.
21858         * config/visium/visium.c: Likewise.
21859         * config/xtensa/xtensa.c: Likewise.
21861 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21863         PR target/81593
21864         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
21865         constraints since the -mupper-regs-* switches have been
21866         eliminated.
21867         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
21868         into a vector from a double word element that was extracted from
21869         another vector, and eliminate extra XXPERMDI instructions.
21870         (vsx_concat_<mode>_2): Likewise.
21871         (vsx_concat_<mode>_3): Likewise.
21872         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
21873         concat to allow optimizing inserts from previous extracts.
21875 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
21877         * config/i386/i386.c (ix86_stack_protect_guard): Generate
21878         memory reference to a SSP offset in TLS address space.
21879         (ix86_print_operand) <case '@'>: Remove.
21880         (ix86_print_operand_punct_valid_p): Remove '@' code.
21881         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
21882         UNSPEC_SP_TLS_TEST.
21883         (stack_tls_protect_set_<mode>): Remove.
21884         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
21885         (stack_tls_protect_test_<mode>): Remove.
21886         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
21888 2017-08-07  Olivier Hainque  <hainque@adacore.com>
21890         PR target/81755
21891         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
21893 2017-08-07  Douglas Rupp  <rupp@adacore.com>
21895         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
21896         variable was referenced as multidir in command.
21898 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21900         PR c/69389
21901         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
21902         BIT_FIELD_REF.
21904 2017-08-07  Martin Liska  <mliska@suse.cz>
21906         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
21907         * config/rl78/rl78.c: Add include of attribs.h.
21908         * config/sh/sh.c: Likewise.
21909         * config/v850/v850.c: Likewise.
21911 2017-08-07  Tom de Vries  <tom@codesourcery.com>
21913         PR middle-end/78266
21914         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
21916 2017-08-07  Martin Liska  <mliska@suse.cz>
21918         * config/mips/mips.c: Include attribs.h.
21920 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21922         PR fortran/68829
21923         * doc/invoke.texi: Document change in behvaior for -Ofast for
21924         Fortran.
21926 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
21928         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
21929         Use gen_frame_mem.
21930         (aarch64_pop_regs): Likewise.
21931         (aarch64_gen_load_pair): Likewise.
21932         (aarch64_save_callee_saves): Likewise.
21933         (aarch64_restore_callee_saves): Likewise.
21935 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21937         * config/i386/i386.c: Revert the last change.
21939 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21941         PR target/81736
21942         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21943         to ...
21944         (ix86_finalize_stack_frame_flags): This.  Also clear
21945         frame_pointer_needed if -fno-omit-frame-pointer is used without
21946         stack access.
21947         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21948         with ix86_finalize_stack_frame_flags.
21949         (ix86_expand_epilogue): Likewise.
21950         (ix86_expand_split_stack_prologue): Likewise.
21952 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21954         PR target/81743
21955         * config/i386/i386.c (get_builtin_code_for_version): Set priority
21956         to P_AES for Westmere.
21958 2017-08-07  Jonathan Yong  <10walls@gmail.com>
21960         * config/i386/mingw.opt (fset-stack-executable): Removed.
21961         * config/i386/cygming.opt (fset-stack-executable): Moved
21962         from mingw.opt.
21963         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
21965 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
21967         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
21969 2017-08-07  Marek Polacek  <polacek@redhat.com>
21971         PR middle-end/81737
21972         * fold-const.c (fold_indirect_ref_1): Check type_domain.
21974 2017-08-07  Martin Liska  <mliska@suse.cz>
21976         * attribs.h (canonicalize_attr_name): New function.
21977         (cmp_attribs): Move from c-format.c and adjusted.
21978         (is_attribute_p): Moved from tree.h.
21979         * tree-inline.c: Add new includes.
21980         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
21981         (private_is_attribute_p): Remove.
21982         (private_lookup_attribute): Likewise.
21983         (private_lookup_attribute_by_prefix): Simplify.
21984         (remove_attribute): Use is_attribute_p.
21985         * tree.h: Remove removed declarations.
21987 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21989         PR middle-end/81698
21990         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
21991         instead of computing it in the function.  Formatting fix.
21992         (expand_case): Don't rely on default_edge being the first edge,
21993         clear it if removing it, pass default_edge to
21994         emit_case_dispatch_table.
21995         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
21996         fix.
21998 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22000         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22001         insn in the function, emit NOP after the insn.
22003 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22005         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22006         and element loops.
22008 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22010         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22011         loop.
22013 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22015         PR tree-optimization/57371
22016         * match.pd: New pattern.
22018 2017-08-04  Marek Polacek  <polacek@redhat.com>
22020         PR middle-end/81695
22021         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22022         perform the computation in offset_int.
22024 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22026         PR tree-optimization/81136
22027         * tree-vectorizer.h: Include tree-hash-traits.h.
22028         (vec_base_alignments): New typedef.
22029         (vec_info): Add a base_alignments field.
22030         (vect_record_base_alignments): Declare.
22031         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22032         field.
22033         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22034         (create_data_ref): Add an is_conditional_in_stmt argument.
22035         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22036         the is_conditional_in_stmt field.
22037         (data_ref_loc): Add an is_conditional_in_stmt field.
22038         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22039         (find_data_references_in_stmt): Update call to create_data_ref.
22040         (graphite_find_data_references_in_stmt): Likewise.
22041         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22042         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22043         (vect_record_base_alignment): New function.
22044         (vect_record_base_alignments): Likewise.
22045         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22046         for nested statements even if we fail to compute a misalignment.
22047         Use pooled base alignments for unconditional references.
22048         (vect_find_same_alignment_drs): Compare base addresses instead
22049         of base objects.
22050         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22051         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22053 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22055         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22056         Add an explicit name for the enum.  Use auto_vec for slp_instances
22057         and grouped_stores.
22058         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22059         for all vectors.
22060         (_bb_vec_info): Add a constructor and destructor.
22061         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22062         (destroy_loop_vec_info): Delete.
22063         (vect_destroy_datarefs): Likewise.
22064         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22065         (vec_info::vec_info): New function.
22066         (vec_info::~vec_info): Likewise.
22067         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22068         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22069         destroy_loop_vec_info.
22070         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22071         (_loop_vec_info::_loop_vec_info): ...this.
22072         (destroy_loop_vec_info): Replace with...
22073         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22074         the stmt_vec_infos.  Leave handling of vec_info information to its
22075         destructor.  Remove explicit vector releases.
22076         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22077         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22078         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22079         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22080         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22081         (destroy_bb_vec_info): Replace with...
22082         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22083         information to its destructor.
22084         (vect_slp_analyze_bb_1): Use new and delete instead of
22085         new_bb_vec_info and destroy_bb_vec_info.
22086         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22087         single delete.
22089 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22091         * tree-data-ref.h (subscript): Add access_fn field.
22092         (data_dependence_relation): Add could_be_independent_p.
22093         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22094         (same_access_functions): Move to tree-data-ref.c.
22095         * tree-data-ref.c (ref_contains_union_access_p): New function.
22096         (access_fn_component_p): Likewise.
22097         (access_fn_components_comparable_p): Likewise.
22098         (dr_analyze_indices): Add a reference to access_fn_component_p.
22099         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22100         DR_ACCESS_FN.
22101         (constant_access_functions): Likewise.
22102         (add_other_self_distances): Likewise.
22103         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22104         (initialize_data_dependence_relation): Use XCNEW and remove
22105         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22106         of access functions that have the same type.  Allow the
22107         subsequence to end with different bases in some circumstances.
22108         Record the chosen access functions in SUB_ACCESS_FN.
22109         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22110         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22111         (subscript_dependence_tester_1): Likewise dra and drb.
22112         (build_classic_dist_vector): Update calls accordingly.
22113         (subscript_dependence_tester): Likewise.
22114         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22115         DDR_COULD_BE_INDEPENDENT_P.
22116         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22117         comp_alias_ddrs instead of may_alias_ddrs.
22118         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22119         New function.
22120         (vect_analyze_data_ref_dependence): Use it if
22121         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22122         distance vectors if that fails.
22123         (dependence_distance_ge_vf): New function.
22124         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22125         LOOP_VINFO_MAY_ALIAS_DDRS.
22127 2017-08-04  Richard Biener  <rguenther@suse.de>
22129         PR middle-end/81705
22130         * fold-const.c (fold_binary_loc): Properly restrict
22131         minus_var0 && minus_var1 case when associating undefined overflow
22132         entities.
22134 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22136         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22138 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22140         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22141         Don't start diagnostic messages with a capital letter.
22142         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22143         Likewise.
22144         (rs6000_invalid_builtin): Likewise.
22145         (rs6000_trampoline_init): Likewise.
22147 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22149         PR target/81621
22150         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22151         after setting changeable df flags.
22153 2017-08-03  Richard Biener  <rguenther@suse.de>
22155         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22156         up if the use is in USE - X.
22158 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22160         * toplev.c (dumpfile.h): New include.
22161         (internal_error_reentered): New static function.  Use it...
22162         (internal_error_function): ...here to handle reentered internal_error.
22164 2017-08-03  Richard Biener  <rguenther@suse.de>
22166         PR middle-end/81148
22167         * fold-const.c (split_tree): Add minus_var and minus_con
22168         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22169         here but always use minus_*.
22170         (associate_trees): Assert we never associate with MINUS_EXPR
22171         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22172         when associating as MINUS_EXPR either.
22173         (fold_binary_loc): Track minus_var and minus_con.
22175 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22177         PR lto/81430
22178         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22179         ACCEL_COMPILER, apply finish_options on
22180         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22182 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22184         PR target/81662
22185         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22186         function_entry_patch_area_size > 0.
22188 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22190         PR driver/81650
22191         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22192         instead of 10??LU, perform unit multiplication in wide_int,
22193         don't change alloc_object_size_limit if the limit is larger
22194         than SSIZE_MAX.
22196         PR tree-optimization/81655
22197         PR tree-optimization/81588
22198         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22199         the case when ranges[i].low and high are 1 for unsigned type with
22200         precision 1.
22202         PR middle-end/81052
22203         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22204         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22206 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22208         * tree-vrp.h: Add include guard.
22210 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22212         PR target/81644
22213         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22214         (ud2): New insn pattern.
22215         * config/i386/i386.c (ix86_expand_epilogue):
22216         For naked functions, generate ud2 instead of trap insn.
22218 2017-08-02  Marek Polacek  <polacek@redhat.com>
22220         PR other/81667
22221         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22223 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22224             Cesar Philippidis  <cesar@codesourcery.com>
22226         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22227         Add missing edge probabilities.
22229 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22231         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22232         Correct endianness.
22234 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22236         PR middle-end/79499
22237         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22238         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22239         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22240         prologue_seq sequences - if any.
22242 2017-08-02  Richard Biener  <rguenther@suse.de>
22244         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22245         via vectors if supported, integer extracts via punning if supported
22246         or otherwise vector extracts.
22248 2017-08-02  Richard Biener  <rguenther@suse.de>
22250         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22251         into ...
22252         (bitmap_insert_into_set): ... this.
22254 2017-08-02  Richard Biener  <rguenther@suse.de>
22256         PR tree-optimization/81633
22257         Revert
22258         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22260         PR tree-optimization/71752
22261         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22263 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22265         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22266         (machine_function::call_ms2sysv_pad_out): Remove field.
22267         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22268         (ix86_compute_frame_layout): Likewise.
22270 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22272         PR target/81654
22273         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22274         attribute with interrupt attribute.
22276 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22278         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22279         BIT_INSERT_EXPR's operand 1
22280         to see if the types precision matches.
22282 2017-08-01  Martin Liska  <mliska@suse.cz>
22284         PR middle-end/70140
22285         * builtins.c (expand_builtin_memcpy_args): Remove.
22286         (expand_builtin_memcpy): Call newly added function
22287         expand_builtin_memory_copy_args.
22288         (expand_builtin_memcpy_with_bounds): Likewise.
22289         (expand_builtin_mempcpy): Remove last argument.
22290         (expand_builtin_mempcpy_with_bounds): Likewise.
22291         (expand_builtin_memory_copy_args): New function created from
22292         expand_builtin_mempcpy_args with small modifications.
22293         (expand_builtin_mempcpy_args): Remove.
22294         (expand_builtin_stpcpy): Remove unused argument.
22295         (expand_builtin): Likewise.
22296         (expand_builtin_with_bounds): Likewise.
22298 2017-08-01  Martin Liska  <mliska@suse.cz>
22300         Revert r250771
22301         Make mempcpy more optimal (PR middle-end/70140).
22303 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22305         PR target/81622
22306         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22307         __builtin_vec_cmpne verify both arguments are compatible vectors
22308         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22309         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22310         move computation of aligned to after checking the argument types.
22311         Formatting fixes.
22313         PR target/80846
22314         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22315         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22316         calls.
22318 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22319             Doug Rupp  <rupp@adacore.com>
22320             Olivier Hainque  <hainque@adacore.com>
22322         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22323         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22324         arm8 (arch v4).
22325         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22326         for TARGET_OS_CPP_BUILTIN.
22327         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22328         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22329         arm_arch7.
22330         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22331         passing required abi options to the assembler for EABI configurations.
22332         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22333         of .text.hot and .text.unlikely sections for kernel modules when
22334         using ARM style exceptions.
22335         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22336         options. Add EXTRA_CC1_SPEC.
22337         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22338         toolchain options.
22339         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22340         transition.
22341         (ARM_TARGET2_DWARF_FORMAT): Define.
22342         * config/arm/t-vxworks: Adjust multilib control to removal of the
22343         Diab command line options.
22345 2017-08-01  Martin Liska  <mliska@suse.cz>
22347         PR gcov-profile/81561
22348         * gcov.c (unblock): Make unblocking safe as we need to preserve
22349         index correspondence of blocks and block_lists.
22351 2017-08-01  Richard Biener  <rguenther@suse.de>
22353         PR tree-optimization/81181
22354         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22355         (compute_antic): ... end of iteration here.
22357 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22359         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22360         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22361         (ftree-slp-vectorize): Likewise.
22362         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22363         can no longer be set independent of flag_tree_loop_vectorize.
22364         * omp-general.c (emp_max_vf): Likewise.
22365         * opts.c (enable_fdo_optimizations): Remove references to
22366         flag_tree_vectorize, these are now implicit.
22367         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22368         and leave it for the options machinery.
22370 2017-08-01  Martin Liska  <mliska@suse.cz>
22372         PR middle-end/70140
22373         * builtins.c (expand_builtin_memcpy_args): Remove.
22374         (expand_builtin_memcpy): Call newly added function
22375         expand_builtin_memory_copy_args.
22376         (expand_builtin_memcpy_with_bounds): Likewise.
22377         (expand_builtin_mempcpy): Remove last argument.
22378         (expand_builtin_mempcpy_with_bounds): Likewise.
22379         (expand_builtin_memory_copy_args): New function created from
22380         expand_builtin_mempcpy_args with small modifications.
22381         (expand_builtin_mempcpy_args): Remove.
22382         (expand_builtin_stpcpy): Remove unused argument.
22383         (expand_builtin): Likewise.
22384         (expand_builtin_with_bounds): Likewise.
22386 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22388         PR target/81641
22389         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22390         print "ds:" only for immediates in generic address space.
22392 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22394         PR target/81639
22395         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22396         (ix86_function_ok_for_sibcall): Return false for naked functions.
22398 2017-08-01  Richard Biener  <rguenther@suse.de>
22400         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22401         (compute_antic): Seed worklist with exit block predecessors.
22402         * cfganal.c (dfs_find_deadend): For a cycle return the source
22403         of the edge closing it.
22405 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22407         * config/aarch64/aarch64.c
22408         (aarch64_can_const_movi_rtx_p): Move 0 check.
22410 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22412         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22413         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22414         above macro.
22415         * match.pd: Ditto in address comparison pattern.
22417 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22419         PR tree-optimization/81627
22420         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22421         closed ssa form for store-store chain.
22423 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22425         PR tree-optimization/81620
22426         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22427         for store-store chain.
22429 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22431         PR tree-optimization/81588
22432         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22433         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22434         swap rhs1 and rhs2 and comparison code unconditionally,
22435         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22436         ranges[i].in_p, instead invert comparison code ccode if
22437         opcode or oe->rank is BIT_IOR_EXPR.
22439         PR target/80846
22440         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22441         a direct optab to conversion optab.
22442         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22443         with GET_MODE_INNER as last argument instead of optab_handler.
22444         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22445         vector extraction if possible and optab is available.
22446         * expr.c (store_constructor): Use convert_optab_handler instead
22447         of optab_handler.  Use vector initialization from smaller
22448         vectors if possible and optab is available.
22449         * tree-vect-stmts.c (vectorizable_load): Likewise.
22450         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22451         now have two modes.
22452         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22453         of vec_init from half-sized vectors with the same element mode.
22454         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22455         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22456         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22457         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22458         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22459         after mode in gen_vec_extract* calls.
22460         (vec_extract<mode>): Renamed to ...
22461         (vec_extract<mode><ssescalarmodelower>): ... this.
22462         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22463         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22464         element mode after mode in gen_vec_init* calls.
22465         (VEC_INIT_HALF_MODE): New mode iterator.
22466         (vec_init<mode>): Renamed to ...
22467         (vec_init<mode><ssescalarmodelower>): ... this.
22468         (vec_init<mode><ssehalfvecmodelower>): New expander.
22469         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22470         (vec_extractv2sfsf): ... this.
22471         (vec_initv2sf): Renamed to ...
22472         (vec_initv2sfsf): ... this.
22473         (vec_extractv2si): Renamed to ...
22474         (vec_extractv2sisi): ... this.
22475         (vec_initv2si): Renamed to ...
22476         (vec_initv2sisi): ... this.
22477         (vec_extractv4hi): Renamed to ...
22478         (vec_extractv4hihi): ... this.
22479         (vec_initv4hi): Renamed to ...
22480         (vec_initv4hihi): ... this.
22481         (vec_extractv8qi): Renamed to ...
22482         (vec_extractv8qiqi): ... this.
22483         (vec_initv8qi): Renamed to ...
22484         (vec_initv8qiqi): ... this.
22485         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22486         (vec_init<mode>): Renamed to ...
22487         (vec_init<mode><VEC_base_l>): ... this.
22488         (vec_extract<mode>): Renamed to ...
22489         (vec_extract<mode><VEC_base_l>): ... this.
22490         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22491         (vec_initv2sfsf): ... this.
22492         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22493         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22494         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22495         element mode after mode in gen_vec_init* calls.
22496         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22497         (vec_init<mode><Vel>): ... this.
22498         (vec_extract<mode>): Renamed to ...
22499         (vec_extract<mode><Vel>): ... this.
22500         * config/aarch64/iterators.md (Vel): New mode attribute.
22501         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22502         Add element mode after mode in gen_vec_extract* calls.
22503         * config/s390/vector.md (non_vec_l): New mode attribute.
22504         (vec_extract<mode>): Renamed to ...
22505         (vec_extract<mode><non_vec_l>): ... this.
22506         (vec_init<mode>): Renamed to ...
22507         (vec_init<mode><non_vec_l>): ... this.
22508         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22509         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22510         vec_extract mode.
22511         * config/arm/iterators.md (V_elem_l): New mode attribute.
22512         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22513         (vec_extract<mode><V_elem_l>): ... this.
22514         (vec_extractv2di): Renamed to ...
22515         (vec_extractv2didi): ... this.
22516         (vec_init<mode>): Renamed to ...
22517         (vec_init<mode><V_elem_l>): ... this.
22518         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22519         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22520         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22521         Add element mode after gen_vec_extract* calls.
22522         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22523         (vec_init<mode><unitmode>): ... this.
22524         (vec_extract<mode>): Renamed to ...
22525         (vec_extract<mode><unitmode>): ... this.
22526         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22527         (vec_init<mode><unitmode>): ... this.
22528         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22529         (vec_initv2sfsf): ... this.
22530         (vec_extractv2sf): Renamed to ...
22531         (vec_extractv2sfsf): ... this.
22532         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22533         Add element mode after gen_vec_extract* calls.
22534         * config/mips/mips.md (unitmode): New mode iterator.
22535         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22536         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22537         * config/spu/spu.md (inner_l): New mode attribute.
22538         (vec_init<mode>): Renamed to ...
22539         (vec_init<mode><inner_l>): ... this.
22540         (vec_extract<mode>): Renamed to ...
22541         (vec_extract<mode><inner_l>): ... this.
22542         * config/sparc/sparc.md (veltmode): New mode iterator.
22543         (vec_init<VMALL:mode>): Renamed to ...
22544         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22545         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22546         (vec_initv2sisi): ... this.
22547         (vec_initv2sf): Renamed to ...
22548         (vec_initv2sfsf): ... this.
22549         (vec_extractv2sf): Renamed to ...
22550         (vec_extractv2sfsf): ... this.
22551         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22552         (vec_init<mode>): Renamed to ...
22553         (vec_init<mode><VEC_base_l>): ... this.
22554         (vec_extract<mode>): Renamed to ...
22555         (vec_extract<mode><VEC_base_l>): ... this.
22556         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22557         (vec_initv2sfsf): ... this.
22558         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22559         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22560         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22561         gen_vec_init* calls.
22563 2017-08-01  Richard Biener  <rguenther@suse.de>
22565         PR tree-optimization/81297
22566         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22567         TREE_OVERFLOW from INTEGER_CSTs.
22569 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22571         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22573 2017-07-31  Carl Love  <cel@us.ibm.com>
22575         * config/rs6000/rs6000-c: Add support for built-in functions
22576         vector signed char vec_xl_be (signed long long, signed char *);
22577         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22578         vector signed int vec_xl_be (signed long long, signed int *);
22579         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22580         vector signed long long vec_xl_be (signed long long, signed long long *);
22581         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22582         vector signed short vec_xl_be (signed long long, signed short *);
22583         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22584         vector double vec_xl_be (signed long long, double *);
22585         vector float vec_xl_be (signed long long, float *);
22586         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22587         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22588         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22589         for the builtins.
22590         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22591         (altivec_expand_builtin): Add switch statement to call
22592         altivec_expand_xl_be for each builtin.
22593         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22594         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22595         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22596         __builtin_vsx_le_be_v16qi.
22597         * doc/extend.texi: Update the built-in documentation file for the
22598         new built-in functions.
22600 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22602         PR target/25967
22603         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22604         New function.
22605         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22607 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22609         * config.gcc: Add z14.
22610         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22611         CPU model numbers for z13s and z14.
22612         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22613         arch12 with z14.
22614         * config/s390/s390-opts.h (enum processor_type): Rename
22615         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22616         * config/s390/s390.c (processor_table): Add field for CPU name to
22617         be passed to Binutils.
22618         (s390_asm_output_machine_for_arch): Use the new field in
22619         processor_table for Binutils.
22620         (s390_expand_builtin): Replace arch12 with z14.
22621         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22622         (s390_get_sched_attrmask): Likewise.
22623         (s390_get_unit_mask): Likewise.
22624         * config/s390/s390.opt: Add z14 to processor_type enum.
22626 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22628         PR hsa/81477
22629         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22630         regardless of optimization level.
22632 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22633             Martin Liska  <mliska@suse.cz>
22635         * predict.def: Remove old comment and adjust probability.
22636         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22637         PREDICT statements.
22639 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22641         PR target/25967
22642         * config/i386/i386.c (ix86_function_naked): New function.
22643         (ix86_can_use_return_insn_p): Return false for naked functions.
22644         (ix86_expand_prologue): Skip prologue for naked functions.
22645         (ix86_expand_epilogue): Skip epilogue for naked functions
22646         and emit trap instruction.
22647         (ix86_warn_func_return): New function.
22648         (ix86_attribute_table): Add "naked" attribute specification.
22649         (TARGET_WARN_FUNC_RETURN): Define.
22650         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22652 2017-07-31  Martin Liska  <mliska@suse.cz>
22654         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22655         (dump_gimple_bb_header): Always dump BB info.
22656         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22658 2017-07-31  Martin Liska  <mliska@suse.cz>
22660         PR sanitize/81530
22661         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22662         also with current_function_decl non-null equality.
22664 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22666         PR sanitizer/81604
22667         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22668         change type to the element type, instead add eltype variable and
22669         use it where we are interested in the element type.
22671         PR tree-optimization/81603
22672         * ipa-polymorphic-call.c
22673         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22674         offset arithmetic in offset_int, bail out if the resulting bit offset
22675         doesn't fit into shwi.
22677 2017-07-31  Martin Liska  <mliska@suse.cz>
22679         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22680         (gimplify_save_expr): Fix comment.
22682 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22684         PR target/79793
22685         * config/i386/i386.c (ix86_function_arg): Update arguments for
22686         exception handler.
22687         (ix86_compute_frame_layout): Set the initial stack offset to
22688         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22689         INCOMING_FRAME_SP_OFFSET.
22690         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22691         stack before exception handler returns.
22692         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22693         the 'ERROR_CODE' for exception handler.
22695 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22697         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22698         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22699         (ASM_OUTPUT_REG_POP): Ditto.
22700         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22701         instead of asm_fprintf to output pure string.
22703 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22705         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22706         to imported_module_or_decl hook.
22707         (debug_nothing_tree_tree_tree_bool): Remove.
22708         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22709         * debug.c (do_nothing_debug_hooks): Use
22710         debug_nothing_tree_tree_tree_bool_bool instead of
22711         debug_nothing_tree_tree_tree_bool.
22712         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22713         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22714         * sdbout.c (sdb_debug_hooks): Likewise.
22715         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22716         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22717         langhook wants it.
22718         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22719         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22720         attribute, don't add anything.
22722 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22724         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22725         (fold_build2_stat_loc): Likewise.
22726         (fold_build3_stat_loc): Likewise.
22727         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22728         (fold_build1_loc): Remove macro.
22729         (fold_build2_loc): Likewise.
22730         (fold_build3_loc): Likewise.
22732 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22734         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22735         (gimple_build_debug_bind_source_stat): Likewise.
22736         * gimple.h (gimple_build_debug_bind): Remove macro.
22737         (gimple_build_debug_bind_source): Likewise.
22739 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22741         * bitmap.c (bitmap_alloc): Adjust.
22742         (bitmap_gc_alloc): Likewise.
22743         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22745 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22747         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22748         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22749         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22750         (bitmap_gc_alloc_stat): Likewise.
22751         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22753 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22755         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22756         * rtl.h (shallow_copy_rtx): Remove macro.
22758 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22760         * emit-rtl.c (gen_raw_REG): Adjust.
22761         * gengenrtl.c (gendef): Likewise.
22762         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22763         * rtl.h (rtx_alloc): Remove macro.
22765 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22767         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22768         (build_tree_list_stat): Likewise.
22769         * tree.h (build_tree_list): Remove macro.
22770         (build_tree_list_vec): Likewise.
22772 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22774         * tree.c (make_vector_stat): Remove _stat from name.
22775         (build_vector_stat): Likewise.
22776         * tree.h (make_vector_stat): Remove macro.
22777         (build_vector_stat): Likewise.
22779 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22781         * tree.h (build_var_debug_value): Remove prototype.
22783 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22785         * tree.c (tree_cons_stat): Remove _stat from name.
22786         * tree.h (tree_cons): Remove macro.
22788 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22790         * tree.c (build_vl_exp_stat): Remove _stat from name.
22791         * tree.h (build_vl_exp): Remove macro.
22793 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22795         * tree.c (build_decl_stat): Remove _stat from name.
22796         * tree.h (build_decl): Remove macro.
22798 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22800         * gimple.c (gimple_build_with_ops_stat): Adjust.
22801         (gimple_alloc_stat): Remove _stat from name.
22802         * gimple.h (gimple_alloc): Remove macro.
22804 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22806         * tree.c (make_tree_vec_stat): Remove _stat from name.
22807         (grow_tree_vec_stat): Likewise.
22808         * tree.h (make_tree_vec_stat): Adjust prototype.
22809         (grow_tree_vec_stat): Likewise.
22810         (make_tree_vec): Remove macro.
22811         (grow_tree_vec): Likewise.
22813 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22815         * fold-const.c (fold_build1_stat_loc): Adjust.
22816         (fold_build2_stat_loc): Likewise.
22817         (fold_build3_stat_loc): Likewise.
22818         * tree.c (build0_stat): Remove _stat from name.
22819         (build1_stat): Likewise.
22820         (build2_stat): Likewise.
22821         (build3_stat): Likewise.
22822         (build4_stat): Likewise.
22823         (build5_stat): Likewise.
22824         * tree.h (build1_loc): Remove macro, and rename _stat function
22825         to this.
22826         (build2_loc): Likewise.
22827         (build3_loc): Likewise.
22828         (build4_loc): Likewise.
22829         (build5_loc): Likewise.
22831 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22833         * tree.c (make_int_cst_stat): Remove _stat from name.
22834         * tree.h (make_int_cst_stat): Adjust prototype.
22835         (make_int_cst): Remove macro.
22837 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22839         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22840         * tree.h (make_tree_binfo_stat): Adjust prototype.
22841         (make_tree_binfo): Remove.
22843 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22845         * tree.c (copy_node_stat): Rename to copy_node.
22846         (build_distinct_type_copy): Adjust.
22847         * tree.h (copy_node_stat): Adjust prototype.
22848         (copy_node): Remove macro.
22850 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22852         * tree.c (make_node_stat): rename to make_node.
22853         (build_tree_list_stat): Adjust.
22854         (build0_stat): Likewise.
22855         (build2_stat): Likewise.
22856         (build3_stat): Likewise.
22857         (build4_stat): Likewise.
22858         (build5_stat): Likewise.
22859         (build_decl_stat): Likewise.
22860         * tree.h (make_node_stat): Adjust prototype.
22861         (make_node): remove macro.
22863 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
22865         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
22866         (PPC_FEATURE2_SCV): Likewise.
22867         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
22869 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22871         * config/aarch64/aarch64.c
22872         (aarch64_internal_mov_immediate): Add new special pattern.
22873         * config/aarch64/aarch64.md (*movdi_aarch64):
22874         Add reg/32bit const mov case.
22876 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22877             Richard Sandiford <richard.sandiford@linaro.org>
22879         * config/aarch64/aarch64.md (mov<mode>): Generalize.
22880         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
22881         Add integer and movi cases.
22882         (movi-split-hf-df-sf split, fp16): New.
22883         (enabled): Added TARGET_FP_F16INST.
22884         * config/aarch64/iterators.md (GPF_HF): New.
22885         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
22887 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22889         * config/aarch64/aarch64.c
22890         (aarch64_simd_container_mode): Add prototype.
22891         (aarch64_expand_mov_immediate): Add HI support.
22892         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
22893         (aarch64_can_const_movi_rtx_p): New.
22894         (aarch64_preferred_reload_class):
22895         Remove restrictions of using FP registers for certain SIMD operations.
22896         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
22897         (aarch64_valid_floating_const): Add integer move validation.
22898         (aarch64_simd_imm_scalar_p): Remove.
22899         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
22900         (aarch64_legitimate_constant_p): Expand list of supported cases.
22901         * config/aarch64/aarch64-protos.h
22902         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
22903         (aarch64_reinterpret_float_as_int): New.
22904         (aarch64_simd_imm_scalar_p): Remove.
22905         * config/aarch64/constraints.md (Uvi): New.
22906         (Dd): Split into Ds and new Dd.
22907         * config/aarch64/aarch64.md (*movsi_aarch64):
22908         Add SIMD mov case.
22909         (*movdi_aarch64): Add SIMD mov case.
22911 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22913         * tree-predcom.c: (struct chain): Handle store-store chain in which
22914         stores for elimination only store loop invariant values.
22915         (execute_pred_commoning_chain): Ditto.
22916         (prepare_initializers_chain_store_elim): Ditto.
22917         (prepare_finalizers): Ditto.
22918         (is_inv_store_elimination_chain): New function.
22919         (initialize_root_vars_store_elim_1): New function.
22921 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22923         * tree-predcom.c: Revise general description of the pass.
22924         (enum chain_type): New enum type for store elimination.
22925         (struct chain): New field supporting store elimination.
22926         (struct component): Ditto.
22927         (dump_chain): Dump store-stores chain.
22928         (release_chain): Release resources.
22929         (split_data_refs_to_components): Compute and create component
22930         contains only stores for elimination.
22931         (get_chain_last_ref_at): New function.
22932         (make_invariant_chain): Initialization.
22933         (make_rooted_chain): Specify chain type in parameter and record it.
22934         (add_looparound_copies): Skip for store-stores chain.
22935         (determine_roots_comp): Compute type of chain and pass it to
22936         make_rooted_chain.
22937         (initialize_root_vars_store_elim_2): New function.
22938         (finalize_eliminated_stores): New function.
22939         (remove_stmt): Handle store for elimination.
22940         (execute_pred_commoning_chain): Execute predictive commoning on
22941         store-store chains.
22942         (determine_unroll_factor): Skip unroll for store-stores chain.
22943         (prepare_initializers_chain_store_elim): New function.
22944         (prepare_initializers_chain): Hanlde store-store chain.
22945         (prepare_finalizers_chain, prepare_finalizers): New function.
22946         (tree_predictive_commoning_loop): Return integer value indicating
22947         if loop is unrolled or lcssa form is corrupted.
22948         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
22950 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22952         * tree-predcom.c (initialize_root): Delete.
22953         (execute_pred_commoning_chain): Initialize root vars and replace
22954         reference of non-combined chain directly, rather than call above
22955         function.
22957 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22959         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
22960         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
22962 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22964         * tree-predcom.c (struct chain): New field init_seq.
22965         (release_chain): Release init_seq.
22966         (prepare_initializers_chain): Record intialization stmts in above
22967         field.
22968         (insert_init_seqs): New function.
22969         (tree_predictive_commoning_loop): Call insert_init_seqs.
22971 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22973         * tree-predcom.c (determine_roots_comp): Skip trivial components.
22975 2017-07-28  Richard Biener  <rguenther@suse.de>
22977         * match.pd: Remove superfluous :c.
22978         * genmatch.c (simplify::id): Add member.
22979         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
22980         Copy id.
22981         (current_id): New global.
22982         (dt_node::parent): Move from ...
22983         (dt_operand::parent): ... here.  Add for_id member.
22984         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
22985         (decision_tree::find_node): Relax order requirement when
22986         merging DT_TRUE nodes to ones inbetween the current simplify
22987         and the one we try to merge with.  Add diagnostic whenever
22988         we need to enforce pattern order by not merging.
22989         (decision_tree::insert): Set current_id.
22990         (decision_tree::print_node): Dump parent node and for_id.
22991         (parser::last_id): Add member.
22992         (parser::push_simplify): Assign unique id.
22993         (parser::parser): Initialize last_id.
22995 2017-07-28  Martin Liska  <mliska@suse.cz>
22997         PR sanitizer/81340
22998         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
22999         gimple_build_debug_bind.
23001 2017-07-28  Richard Biener  <rguenther@suse.de>
23003         PR tree-optimization/81502
23004         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23005         BIT_FIELD_REF.
23006         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23007         size/pos operands.
23008         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23009         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23010         for BIT_FIELD_REF args.
23011         * fold-const.c (make_bit_field_ref): Likewise.
23012         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23014 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23016         PR sanitizer/80998
23017         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23018         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23019         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23020         Or it into SANITIZER_UNDEFINED.
23021         * ubsan.c: Include gimple-fold.h and varasm.h.
23022         (ubsan_expand_ptr_ifn): New function.
23023         (instrument_pointer_overflow): New function.
23024         (maybe_instrument_pointer_overflow): New function.
23025         (instrument_object_size): Formatting fix.
23026         (pass_ubsan::execute): Call instrument_pointer_overflow
23027         and maybe_instrument_pointer_overflow.
23028         * internal-fn.c (expand_UBSAN_PTR): New function.
23029         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23030         * sanitizer.def (__ubsan_handle_pointer_overflow,
23031         __ubsan_handle_pointer_overflow_abort): New builtins.
23032         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23033         * internal-fn.def (UBSAN_PTR): New internal function.
23034         * opts.c (sanitizer_opts): Add pointer-overflow.
23035         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23036         * fold-const.c (build_range_check): Compute pointer range check in
23037         integral type if pointer arithmetics would be needed.  Formatting
23038         fixes.
23040 2017-07-28  Martin Liska  <mliska@suse.cz>
23042         PR sanitizer/81460
23043         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23044         parameters that are of a variable-length.
23046 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23048         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23049         rs6000/biarch64.h.
23050         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23051         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23052         (CRT_CALL_STATIC_FUNCTION): Likewise.
23053         (ASM_DEFAULT_SPEC): New define.
23054         (ASM_SPEC32): Likewise.
23055         (ASM_SPEC64): Likewise.
23056         (ASM_SPEC_COMMON): Likewise.
23057         (ASM_SPEC): Likewise.
23058         (INVALID_64BIT): Likewise.
23059         (LINK_OS_DEFAULT_SPEC): Likewise.
23060         (LINK_OS_SPEC32): Likewise.
23061         (LINK_OS_SPEC64): Likewise.
23062         (POWERPC_LINUX): Likewise.
23063         (PTRDIFF_TYPE): Likewise.
23064         (RESTORE_FP_PREFIX): Likewise.
23065         (RESTORE_FP_SUFFIX): Likewise.
23066         (SAVE_FP_PREFIX): Likewise.
23067         (SAVE_FP_SUFFIX): Likewise.
23068         (SIZE_TYPE): Likewise.
23069         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23070         (TARGET_64BIT): Likewise.
23071         (TARGET_64BIT): Likewise.
23072         (TARGET_AIX): Likewise.
23073         (WCHAR_TYPE_SIZE): Likewise.
23074         (WCHAR_TYPE): Undefine.
23075         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23076         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23077         (CPP_OS_RTEMS_SPEC): Delete.
23078         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23079         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23080         link_os_spec64.
23081         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23083 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23085         PR tree-optimization/81578
23086         * tree-parloops.c (build_new_reduction): Bail out if
23087         reduction_code isn't one of the standard OpenMP reductions.
23088         Move the details printing after that decision.
23090 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23092         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23093         related to reload_in_progress.
23094         (splat_input_operand): Likewise.
23095         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23096         Delete prototype.
23097         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23098         field.
23099         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23100         (TARGET_INSTANTIATE_DECLS): Likewise.
23101         (legitimate_indexed_address_p): Delete reload_in_progress code.
23102         (rs6000_debug_legitimate_address_p): Likewise.
23103         (rs6000_eliminate_indexed_memrefs): Likewise.
23104         (rs6000_emit_le_vsx_store): Likewise.
23105         (rs6000_emit_move_si_sf_subreg): Likewise.
23106         (rs6000_emit_move): Likewise.
23107         (register_to_reg_type): Likewise.
23108         (rs6000_pre_atomic_barrier): Likewise.
23109         (rs6000_machopic_legitimize_pic_address): Likewise.
23110         (rs6000_allocate_stack_temp): Likewise.
23111         (rs6000_address_for_fpconvert): Likewise.
23112         (rs6000_address_for_altivec): Likewise.
23113         (rs6000_secondary_memory_needed_rtx): Delete function.
23114         (rs6000_check_sdmode): Likewise.
23115         (rs6000_alloc_sdmode_stack_slot): Likewise.
23116         (rs6000_instantiate_decls): Likewise.
23117         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23118         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23119         Delete reload_in_progress.
23120         (*vec_reload_and_plus_<mptrsize>): Likewise.
23121         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23122         (vsx_div_v2di): Likewise.
23123         (vsx_udiv_v2di): Likewise.
23125 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23127         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23128         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23129         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23130         (rs6000_debug_reg_global): Delete print of LRA status.
23131         (rs6000_option_override_internal): Delete dead LRA related code.
23132         (rs6000_lra_p): Delete function.
23133         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23135 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23137         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23138         * config/riscv/rtems.h: New file.
23140 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23141             Sudakshina Das  <sudi.das@arm.com>
23143         * config/aarch64/aarch64.md
23144         (define_split for and<mode>3nr_compare): Move
23145         non aarch64_logical_operand to a register.
23146         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23147         register immediate operand to a register.
23148         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23150 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23152         PR middle-end/81564
23153         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23155 2017-07-27  Richard Biener  <rguenther@suse.de>
23157         PR tree-optimization/81573
23158         PR tree-optimization/81494
23159         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23160         multi defuse cycle case.
23162 2017-07-27  Richard Biener  <rguenther@suse.de>
23164         PR tree-optimization/81571
23165         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23166         PHIs.
23168 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23170         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23171         earlier and only if MASK_FPU is set.  Adjust formatting.
23173 2017-07-27  Martin Liska  <mliska@suse.cz>
23175         * opt-functions.awk: Add validation of value of Init.
23176         * optc-gen.awk: Pass new argument.
23178 2017-07-27  Martin Liska  <mliska@suse.cz>
23180         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23181         Fix wrong condition.
23183 2017-07-27  Martin Liska  <mliska@suse.cz>
23185         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23186         BBs and edges seen by autoFDO.
23188 2017-07-27  Richard Biener  <rguenther@suse.de>
23190         PR tree-optimization/81502
23191         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23192         with incompatible but same sized type.
23193         (execute_update_addresses_taken): Likewise.
23195 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23197         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23198         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23200 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23202         PR target/81534
23203         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23204         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23205         Change s_operand to memory_operand.
23207 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23209         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23210         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23211         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23212         Emit instructions rather than returning an expression.  Handle TFmode
23213         and KFmode by casting to TImode.
23214         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23215         (rs6000_emit_le_vsx_store): Likewise.
23216         * config/rs6000/vsx.md (VSX_TI): New iterator.
23217         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23218         (*vsx_le_undo_permute_<mode>): Likewise.
23219         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23220         emit the split sequence.
23221         (*vsx_le_perm_store_<mode>): Likewise.
23223 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23225         PR tree-optimization/81555
23226         PR tree-optimization/81556
23227         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23228         if true, force CHANGED for the recursive invocation.
23229         (reassociate_bb): Remember original length of ops array, pass
23230         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23232         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23233         attributes for noipa attribute.  For naked attribute use
23234         lookup_attribute first before lookup_attribute_spec.
23235         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23236         noipa attribute.
23237         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23238         for functions with noipa attribute.
23239         (cgraph_externally_visible_p): Return true for functions with noipa
23240         attribute.
23241         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23242         for functions with noipa attribute.
23243         * doc/extend.texi: Document noipa function attribute.
23244         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23245         also for functions with noipa attribute.
23246         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23248 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23250         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23251         vec_unalign_load_cost and vec_unalign_store_cost.
23253 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23255         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23256         -mvsx-small-integer option.
23257         (ISA_3_0_MASKS_IEEE): Likewise.
23258         (OTHER_VSX_VECTOR_MASKS): Likewise.
23259         (POWERPC_MASKS): Likewise.
23260         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23261         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23262         code, only testing for DImode being allowed in non-VSX floating
23263         point registers.
23264         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23265         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23266         another VSX test.
23267         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23268         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23269         TARGET_P8_VECTOR test.
23270         (rs6000_secondary_reload_simple_move): Likewise.
23271         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23272         since TARGET_P9_VECTOR was already tested.
23273         (rs6000_opt_masks): Remove -mvsx-small-integer.
23274         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23275         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23276         used.
23277         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23278         test for TARGET_VEXTRACTUB was used, and that uses
23279         TARGET_P9_VECTOR.
23280         (p9 extract splitter): Likewise.
23281         (vsx_extract_<mode>_di_p9): Likewise.
23282         (vsx_extract_<mode>_store_p9): Likewise.
23283         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23284         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23285         the elimination of TARGET_VSX_SMALL_INTEGER.
23286         (vsx_extract_<mode>_p8): Likewise.
23287         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23288         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23289         (vsx_set_<mode>_p9): Likewise.
23290         (vsx_set_v4sf_p9): Likewise.
23291         (vsx_set_v4sf_p9_zero): Likewise.
23292         (vsx_insert_extract_v4sf_p9): Likewise.
23293         (vsx_insert_extract_v4sf_p9_2): Likewise.
23294         * config/rs6000/rs6000.md (sign extend splitter): Change
23295         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23296         (floatsi<mode>2_lfiwax_mem): Likewise.
23297         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23298         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23299         since a test for TARGET_P9_VECTOR was used.
23300         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23301         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23302         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23303         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23304         TARGET_P8_VECTOR test.
23305         (fix_trunc<mode>si2_stfiwx): Likewise.
23306         (fix_trunc<mode>si2_internal): Likewise.
23307         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23308         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23309         used.
23310         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23311         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23312         TARGET_P8_VECTOR test.
23313         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23314         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23315         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23316         used.
23317         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23318         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23319         since a test for TARGET_P9_VECTOR was used.
23320         (splitter for loading small constants): Likewise.
23322 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23324         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23325         vec_fp_stmt_cost.
23327 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23329         PR target/81563
23330         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23331         (fp_valid_at): Likewise.
23333 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23335         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23336         (qdf24xx_addrcost_table): Likewise.
23337         (cortexa57_tunings): Update to use generic_branch_cost.
23338         (cortexa72_tunings): Likewise.
23339         (cortexa73_tunings): Likewise.
23340         (qdf24xx_tunings): Likewise.
23342 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23344         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23345         (thunderx2t99_branch_cost): Likewise.
23346         (cortexa35_tunings): Update to use generic_branch_cost.
23347         (cortexa53_tunings): Likewise.
23348         (cortexa57_tunings): Likewise.
23349         (cortexa72_tunings): Likewise.
23350         (cortexa73_tunings): Likewise.
23351         (thunderx2t99_tunings): Likewise.
23353 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23355         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23356         (sparc_option_override): Honour MASK_FSMULD.
23357         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23358         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23359         * config/sparc/sparc.opt (mfsmuld): New option.
23360         * doc/invoke.texi (mfsmuld): Document option.
23362 2017-07-26  Marek Polacek  <polacek@redhat.com>
23364         PR middle-end/70992
23365         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23367 2017-07-26  Richard Biener  <rguenther@suse.de>
23369         * gimple-match-head.c (do_valueize): Return OP if valueize
23370         returns NULL_TREE.
23371         (get_def): New helper to get at the def stmt of a SSA name
23372         if valueize allows.
23373         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23374         do_valueize to get at the def stmt.
23375         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23377 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23379         PR middle-end/46932
23380         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23382 2017-07-26  Martin Liska  <mliska@suse.cz>
23384         PR sanitize/81186
23385         * function.c (expand_function_start): Make expansion of
23386         nonlocal_goto_save_area after parm_birth_insn.
23388 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23390         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23391         from all CPU target flags enable members.
23393 2017-07-26  Richard Biener  <rguenther@suse.de>
23395         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23396         (decision_tree::gen): Make 'type' const.
23397         (write_predicate): Likewise.
23399 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23401         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23402         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23403         (rs6000_option_override_internal): Likewise.
23404         (rs6000_expand_vector_set): Likewise.
23405         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23406         (TARGET_UPPER_REGS_SF): Likewise.
23407         (TARGET_UPPER_REGS_DI): Likewise.
23408         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23409         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23410         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23411         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23412         (Splitters for DI constants in Altivec registers): Likewise.
23413         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23414         (vsx_set_v4sf_p9): Likewise.
23415         (vsx_set_v4sf_p9_zero): Likewise.
23416         (vsx_insert_extract_v4sf_p9): Likewise.
23417         (vsx_insert_extract_v4sf_p9_2): Likewise.
23419 2017-07-25  Carl Love  <cel@us.ibm.com>
23421         * doc/extend.texi: Update the built-in documentation file for the
23422         existing built-in functions
23423         vector signed char vec_cnttz (vector signed char);
23424         vector unsigned char vec_cnttz (vector unsigned char);
23425         vector signed short vec_cnttz (vector signed short);
23426         vector unsigned short vec_cnttz (vector unsigned short);
23427         vector signed int vec_cnttz (vector signed int);
23428         vector unsigned int vec_cnttz (vector unsigned int);
23429         vector signed long long vec_cnttz (vector signed long long);
23430         vector unsigned long long vec_cnttz (vector unsigned long long);
23432 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23434         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23435         accesses where the use is for the first operand of a BIT_INSERT.
23437 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23439         PR bootstrap/81521
23440         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23441         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23443 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23445         * config/i386/gstabs.h: Delete.
23446         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23448 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23450         * config/i386/i386.c (ix86_decompose_address): Do not check for
23451         register RTX when looking at index_reg or base_reg.
23452         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23454 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23456         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23457         to update EH info here.
23459 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23461         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23463 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23465         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23467 2017-07-25  Torsten Duwe  <duwe@suse.de>
23469         * common.opt: Introduce -fpatchable-function-entry
23470         command line option, and its variables function_entry_patch_area_size
23471         and function_entry_patch_area_start.
23472         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23473         including a two-value parser.
23474         * target.def (print_patchable_function_entry): New target hook.
23475         * targhooks.h (default_print_patchable_function_entry): New function.
23476         * targhooks.c (default_print_patchable_function_entry): Likewise.
23477         * toplev.c (process_options): Switch off IPA-RA if
23478         patchable function entries are being generated.
23479         * varasm.c (assemble_start_function): Look at the
23480         patchable-function-entry command line switch and current
23481         function attributes and maybe generate NOP instructions by
23482         calling the print_patchable_function_entry hook.
23483         * doc/extend.texi: Document patchable_function_entry attribute.
23484         * doc/invoke.texi: Document -fpatchable_function_entry
23485         command line option.
23486         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23487         New target hook.
23488         * doc/tm.texi: Re-generate.
23490 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23492         PR target/81532
23493         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23494         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23496 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23498         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23500 2017-07-25  Richard Biener  <rguenther@suse.de>
23502         PR tree-optimization/81455
23503         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23504         not walk in cycles when looking for guards.
23506 2017-07-25  Richard Biener  <rguenther@suse.de>
23508         PR tree-optimization/81529
23509         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23510         when optimizing backedge uses.
23512 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23514         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23515         character for AIX.
23516         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23517         to dl_section_ref.  On AIX, append an expression to subtract
23518         the size of the section length to dl_section_ref.
23520 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23522         * configure.ac: If any of the config.* scripts fail, exit 1.
23523         * configure: Regenerate.
23525 2017-07-25  Richard Biener  <rguenther@suse.de>
23527         PR middle-end/81546
23528         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23529         of immediate uses, be more verbose on errors.
23531 2017-07-25  Richard Biener  <rguenther@suse.de>
23533         PR tree-optimization/81510
23534         * tree-vect-loop.c (vect_is_simple_reduction): When the
23535         reduction stmt is not inside the loop bail out.
23537 2017-07-25  Richard Biener  <rguenther@suse.de>
23539         PR tree-optimization/81303
23540         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23541         profitability check against LOOP_VINFO_NITERSM1.
23543 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23545         * domwalk.c (cmp_bb_postorder): Simplify.
23546         (sort_bbs_postorder): New function.  Use it...
23547         (dom_walker::walk): ...here to optimize common cases.
23549 2017-07-25  Martin Liska  <mliska@suse.cz>
23551         PR ipa/81520
23552         * ipa-visibility.c (function_and_variable_visibility): Make the
23553         redirection just on target that supports aliasing.
23554         Fix GNU coding style.
23556 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23558         PR libgcc/61152
23559         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23560         Format changes.
23561         * config/arm/rtems.h: Likewise.
23562         * config/bfin/rtems.h: Likewise.
23563         * config/i386/rtemself.h: Likewise.
23564         * config/lm32/rtems.h: Likewise.
23565         * config/m32c/rtems.h: Likewise.
23566         * config/m68k/rtemself.h: Likewise.
23567         * config/microblaze/rtems.h: Likewise.
23568         * config/mips/rtems.h: Likewise.
23569         * config/moxie/rtems.h: Likewise.
23570         * config/nios2/rtems.h: Likewise.
23571         * config/powerpcspe/rtems.h: Likewise.
23572         * config/rs6000/rtems.h: Likewise.
23573         * config/rtems.h: Likewise.
23574         * config/sh/rtems.h: Likewise.
23575         * config/sh/rtemself.h: Likewise.
23576         * config/sparc/rtemself.h: Likewise.
23578 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23580         PR 81487
23581         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23582         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23583         * tree-ssa-structalias.c (alias_get_name): Same.
23585 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23587         PR target/81414
23588         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23589         instructions if no du chain is found.
23591 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23593         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23595 2017-07-25  Richard Biener  <rguenther@suse.de>
23597         PR middle-end/81505
23598         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23599         sticky.
23601 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23603         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23604         upper-regs options.
23605         (ISA_2_7_MASKS_SERVER): Likewise.
23606         (ISA_3_0_MASKS_IEEE): Likewise.
23607         (OTHER_P8_VECTOR_MASKS): Likewise.
23608         (OTHER_VSX_VECTOR_MASKS): Likewise.
23609         (POWERPC_MASKS): Likewise.
23610         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23611         duplicate list of options.
23612         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23613         explicit -mupper-regs options.
23614         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23615         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23616         alias for -mupper-regs-df.
23617         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23618         (rs6000_init_hard_regno_mode_ok): Likewise.
23619         (rs6000_option_override_internal): Likewise.
23620         (rs6000_opt_masks): Likewise.
23621         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23622         options in terms of whether -mvsx or -mpower8-vector was used.
23623         (TARGET_UPPER_REGS_DI): Likewise.
23624         (TARGET_UPPER_REGS_SF): Likewise.
23625         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23626         -mupper-regs-* options.
23628 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23630         * passes.c (emergency_dump_function): Print some empty lines and a
23631         header before the RTL dump.
23633 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23635         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23637 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23639         PR target/79041
23640         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23641         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23643 2017-07-24  Carl Love  <cel@us.ibm.com>
23645         * config/rs6000/rs6000-c.c: Add support for built-in functions
23646         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23647         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23648         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23649         vec_extract_fp_from_shortl): Add defines for the two builtins.
23650         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23651         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23652         new builtins.
23653         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23654         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23655         * doc/extend.texi: Update the built-in documentation file for the
23656         new built-in function.
23658 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23660         PR bootstrap/81521
23661         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23662         documentation.
23663         * doc/generic.texi: Likewise.
23664         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23665         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23667 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23669         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23670         (aarch64_mls_elt_merge<mode>): Likewise.
23672 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23674         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23675         having __cxa_atexit.
23677 2017-07-23  Michael Collison  <michael.collison@arm.com>
23679         * config/arm/arm.c (arm_option_override): Deprecate
23680         use of -mstructure-size-boundary.
23681         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23682         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23684 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23686         PR target/80695
23687         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23688         Reduce cost estimate for direct moves.
23690 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23692         PR target/80569
23693         * config/i386/i386.c (ix86_option_override_internal): Disable
23694         BMI, BMI2 and TBM instructions for -m16.
23696 2017-07-21  Carl Love  <cel@us.ibm.com>
23698         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23699         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23700         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23701         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23702         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23703         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23704         VMULOSW): New enum "unspec" values.
23705         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23706         altivec_vmulosw): New patterns.
23707         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23708         VMULOSW): Add definitions.
23710 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23712         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23713         (qdf24xx): Likewise.
23714         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23715         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23716         (AARCH64_FL_V8_1): Renumber.
23717         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23718         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23719         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23720         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23721         rdma to feature modifiers list.
23723 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23725         PR middle-end/56727
23726         * ipa-visibility (function_and_variable_visibility): Convert
23727         recursive PLT call to direct call if appropriate.
23729 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23731         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23732         operand 1 to see if the types precision matches.
23733         * fold-const.c (operand_equal_p): Likewise.
23735 2017-07-21  Richard Biener  <rguenther@suse.de>
23737         PR tree-optimization/81303
23738         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23739         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23740         (vect_peeling_hash_get_lowest_cost): Adjust.
23741         (vect_enhance_data_refs_alignment): Likewise.  Use
23742         vect_get_peeling_costs_all_drs to compute the penalty for no
23743         peeling to match up costs.
23745 2017-07-21  Richard Biener  <rguenther@suse.de>
23747         PR tree-optimization/81500
23748         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23749         we didn't identify a reduction path.
23751 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23752             Cesar Philippidis  <cesar@codesourcery.com>
23754         PR gcov-profile/81442
23755         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23756         probabilities.
23758 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23760         PR lto/81430
23761         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23762         function.
23763         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23764         nvptx_override_options_after_change.
23766 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23768         * dwarf2out.c (output_file_names): Avoid double testing for
23769         dwarf_version >= 5.
23771 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23773         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23775 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23777         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23778         hot/cold regions.
23779         (try_crossjump_to_edge): Do not punt on partitioned functions.
23781 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23783         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23784         Put all BBs reachable only via paths crossing cold region to cold
23785         region.
23786         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23788 2016-07-21  Richard Biener  <rguenther@suse.de>
23790         PR tree-optimization/81303
23791         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23792         into account prologue and epilogue iterations when raising
23793         min_profitable_iters to sth at least covering one vector iteration.
23795 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23797         * config/arm/arm.c (arm_test_cpu_arch_dat):
23798         Check for overlap.
23800 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23802         Remove TYPE_METHODS.
23803         * tree.h (TYPE_METHODS): Delete.
23804         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23805         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23806         (dbxout_type_methods): Scan TYPE_FIELDS.
23807         (dbxout_type): Don't check TYPE_METHODS here.
23808         * function.c (use_register_for_decl): Always ignore register for
23809         class types when not optimizing.
23810         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23811         * tree.c (free_lang_data_in_type): Stitch out member functions and
23812         templates from TYPE_FIELDS.
23813         (build_distinct_type_copy, verify_type_variant,
23814         verify_type): Member fns are on TYPE_FIELDS.
23815         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23816         * tree-pretty-print.c (dump_generic_node): Likewise.
23818 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23820         PR target/80846
23821         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23822         V2TImode and V4TImode.
23823         (ix86_expand_vector_extract): Likewise.
23824         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23825         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23826         (ssescalarmode): Handle V4TImode and V2TImode.
23827         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23828         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23829         (VEXTRACTI128_MODE): New mode iterator.
23830         (splitter for *vec_extractv?ti first element): New.
23831         (VEC_INIT_MODE): New mode iterator.
23832         (vec_init<mode>): Consolidate 3 expanders into one using
23833         VEC_INIT_MODE mode iterator.
23835 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23837         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23838         non_spilled_static_chain_regno_p.
23840 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23842         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23844 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23846         * bb-reorder.c (connect_traces): Allow copying of blocks within
23847         single partition.
23849 2017-07-20  Richard Biener  <rguenther@suse.de>
23851         * gimple.h (gimple_phi_result): Add gphi * overload.
23852         (gimple_phi_result_ptr): Likewise.
23853         (gimple_phi_arg): Likewise.  Adjust index assert to only
23854         allow actual argument accesses rather than all slots available
23855         by capacity.
23856         (gimple_phi_arg_def): Add gphi * overload.
23857         * tree-phinodes.c (make_phi_node): Initialize only actual
23858         arguments.
23859         (resize_phi_node): Clear memory not covered by old node,
23860         do not initialize excess argument slots.
23861         (reserve_phi_args_for_new_edge): Initialize new argument slot
23862         completely.
23864 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
23866         PR tree-optimization/81388
23867         Revert r238585:
23868         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
23870         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
23871         by removing computation of may_be_zero.
23873 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23874             Tom de Vries  <tom@codesourcery.com>
23876         PR middle-end/81030
23877         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
23878         when gimple level profile disagrees with what RTL expander did.
23880 2017-07-20  Richard Biener  <rguenther@suse.de>
23882         PR tree-optimization/61171
23883         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
23884         (vect_analyze_stmt): Add slp instance parameter.
23885         (vectorizable_reduction): Likewise.
23886         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
23887         (vect_is_simple_reduction): Deal with chains not detected
23888         as SLP reduction chain, specifically not properly associated
23889         chains containing a mix of plus/minus.
23890         (get_reduction_op): Remove.
23891         (get_initial_defs_for_reduction): Simplify, pass in whether
23892         this is a reduction chain, pass in the SLP node for the PHIs.
23893         (vect_create_epilog_for_reduction): Get the SLP instance as
23894         arg and adjust.
23895         (vectorizable_reduction): Get the SLP instance as arg.
23896         During analysis remember the SLP node with the PHIs in the
23897         instance.  Simplify getting at the vectorized reduction PHIs.
23898         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
23899         through SLP instance.
23900         (vect_slp_analyze_operations): Likewise.
23901         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
23902         (vect_transform_stmt): Likewise.
23904 2017-07-20  Tom de Vries  <tom@codesourcery.com>
23906         PR tree-optimization/81489
23907         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
23908         read of phi arg location to before loop that modifies phi.
23910 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
23912         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
23913         New pattern.
23915 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23917         PR middle-end/81331
23918         * except.c (execute): Fix ordering issue.
23920 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23922         PR rtl-optimization/81423
23923         * combine.c (make_compound_operation_int): Don't try to optimize
23924         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
23926 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23928         PR rtl-optimization/81423
23929         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
23930         with a constant that is -1 in the truncated to mode.
23932 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23934         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
23935         (determine_unlikely_bbs): ... here.
23936         * predict.h (propagate_unlikely_bbs_forward): Declare.
23937         * cfgexpand.c (pass_expand::execute): Use it.
23938         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
23939         unlikely edges.
23940         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
23941         propagate_unlikely_bbs_forward.
23943 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23945         PR middle-end/81331
23946         * except.c (maybe_add_nop_after_section_switch): New function.
23947         (execute): Use it.
23949 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23951         * gimple.h (gimple_phi_set_arg): Make assert more strict.
23953 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23955         * gimple.h (gimple_phi_arg): Make assert more strict.
23957 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
23959         * config.gcc (powerpc*-*-*): Add mmintrin.h.
23960         * config/rs6000/mmintrin.h: New file.
23961         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
23963 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
23965         PR tree-optimization/81346
23966         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
23968 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23970         * config/nvptx/nvptx.md (VECIM): Add V2DI.
23972 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23974         * config/nvptx/nvptx-modes.def: Add V2DImode.
23975         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
23976         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
23977         (nvptx_output_mov_insn): Handle lack of mov.b128.
23978         (nvptx_print_operand): Handle 'H' and 'L' codes.
23979         (nvptx_vector_mode_supported): Allow V2DImode.
23980         (nvptx_preferred_simd_mode): New function.
23981         (nvptx_data_alignment): New function.
23982         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
23983         nvptx_preferred_simd_mode.
23984         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
23985         64 to 128 bits.
23986         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
23988 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23990         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
23991         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
23992         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
23993         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
23994         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
23995         (mov<VECIM>_insn): New define_insn.
23996         (define_expand "mov<VECIM>): New define_expand.
23998 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24000         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24002 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24004         PR tree-optimization/81346
24005         * fold-const.h (fold_div_compare, range_check_type): Declare.
24006         * fold-const.c (range_check_type): New function.
24007         (build_range_check): Use range_check_type.
24008         (fold_div_compare): No longer static, rewritten into
24009         a match.pd helper function.
24010         (fold_comparison): Don't call fold_div_compare here.
24011         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24012         as helper function.
24014 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24016         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24017         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24018         * tree.c (find_decls_types_r, verify_type): Use
24019         TYPE_{MIN,MAX}_VALUE_RAW.
24020         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24021         (hash_tree): Likewise.
24022         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24023         Likewise.
24024         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24025         Likewise.
24027 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24029         PR middle-end/81464
24030         * omp-expand.c (expand_omp_for_static_chunk): Handle
24031         equal-argument loop exit phi.
24033 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24035         PR target/81471
24036         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24037         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24038         operand 2 predicate.
24039         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24040         operand 2 predicate.
24041         (ror,rol -> rorx splitters): Use const_int_operand as
24042         operand 2 predicate.
24044 2017-06-18  Richard Biener  <rguenther@suse.de>
24046         PR tree-optimization/81410
24047         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24048         the gap in the ! slp_perm SLP case after each group.
24050 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24052         PR middle-end/81463
24053         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24054         again.
24056 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24058         PR middle-end/81462
24059         * predict.c (set_even_probabilities): Cleanup; do not affect
24060         probabilities that are already known.
24061         (combine_predictions_for_bb): Call even when count is set.
24063 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24065         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24066         TYPE_MAX_VALUE.
24068 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24070         PR target/81408
24071         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24072         optimization for loop niter analysis.
24074 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24076         PR target/81473
24077         * config/avr/avr.c (avr_optimize_casesi): Don't use
24078         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24080 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24082         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24083         body_cost_vec from _vect_peel_extended_info.
24084         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24085         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24086         npeel.
24088 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24090         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24092 2017-07-18  Richard Biener  <rguenther@suse.de>
24094         PR tree-optimization/80620
24095         PR tree-optimization/81403
24096         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24097         info when re-using a VN table entry.
24099 2017-07-18  Richard Biener  <rguenther@suse.de>
24101         PR tree-optimization/81418
24102         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24103         vectype_in.  Verify that with lane-reducing reduction operations
24104         we have a single def-use cycle.
24106 2017-07-17  Carl Love  <cel@us.ibm.com>
24108         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24110         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24111         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24112         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24113         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24114         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24115         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24116         VMULOSW): New enum "unspec" values.
24117         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24118         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24119         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24120         altivec_vmulosw): New patterns.
24121         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24122         VMULOSW): Add definitions.
24124 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24126         * config/alpha/alpha.c: Include predict.h.
24128 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24130         * tree-vrp.c (compare_assert_loc): Fix comparison function
24131         to return predictable results.
24133 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24135         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24136         option.
24137         (subdi3): Likewise.
24138         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24139         * doc/invoke.texi (mexpand-adddi): Update text.
24141 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24143         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24144         that also clobbers the CC register. The old expand code is moved
24145         to ...
24146         (*arc_clzsi2): ... here.
24147         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24148         the CC register. The old expand code is moved to ...
24149         (arc_ctzsi2): ... here.
24151 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24153         * config/arc/arc.opt (mindexed-loads): Use initial value
24154         TARGET_INDEXED_LOADS_DEFAULT.
24155         (mauto-modify-reg): Use initial value
24156         TARGET_AUTO_MODIFY_REG_DEFAULT.
24157         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24158         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24159         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24160         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24162 2017-07-17  Martin Liska  <mliska@suse.cz>
24164         PR sanitizer/81302
24165         * opts.c (finish_options): Do not allow -fgnu-tm
24166         w/ -fsanitize={kernel-,}address.  Say sorry.
24168 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24170         PR target/81369
24171         * tree-loop-distribution.c (classify_partition): Only assert on
24172         numer of iterations.
24173         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24174         (distribute_loop): Remove code handling loop with unknown niters.
24175         (pass_loop_distribution::execute): Skip loop with unknown niters.
24177 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24179         PR target/81369
24180         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24181         function sort_partitions_by_post_order.
24183 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24185         PR tree-optimization/81374
24186         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24187         the max index of basic blocks, rather than number of basic blocks.
24189 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24191         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24192         proto.
24193         (arc_legitimate_pic_operand_p): Likewise.
24194         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24195         function.
24196         (arc_needs_pcl_p): Likewise.
24197         (arc_legitimate_pc_offset_p): Likewise.
24198         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24199         function is also used in constrains.md.
24200         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24201         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24202         PLUS.  Only return true/false in known cases, otherwise assert.
24203         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24204         is already called in arc_legitimate_constant_p.
24205         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24206         pic addresses.
24207         (LEGITIMATE_PIC_OPERAND_P): Use
24208         arc_raw_symbolic_reference_mentioned_p function.
24209         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24210         function.
24211         (Cal): Likewise.
24212         (C32): Likewise.
24214 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24215         Andrew Burgess  <andrew.burgess@embecosm.com>
24217         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24218         (arc_return_address_register): New function.
24219         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24220         (arc_handle_fndecl_attribute): Add naked attribute.
24221         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24222         (TARGET_WARN_FUNC_RETURN): Likewise.
24223         (arc_allocate_stack_slots_for_args): New function.
24224         (arc_warn_func_return): Likewise.
24225         (machine_function): Change type fn_type.
24226         (arc_compute_function_type): Consider new naked function type,
24227         change function return type.
24228         (arc_must_save_register): Adapt to handle new
24229         arc_compute_function_type's return type.
24230         (arc_expand_prologue): Likewise.
24231         (arc_expand_epilogue): Likewise.
24232         (arc_return_address_regs): Delete.
24233         (arc_return_address_register): New function.
24234         (arc_epilogue_uses): Use above function.
24235         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24236         (arc_function_type): Change encoding, add naked type.
24237         (ARC_INTERRUPT_P): Change to handle the new encoding.
24238         (ARC_FAST_INTERRUPT_P): Likewise.
24239         (ARC_NORMAL_P): Define.
24240         (ARC_NAKED_P): Likewise.
24241         (arc_compute_function_type): Delete prototype.
24242         * config/arc/arc.md (in_ret_delay_slot): Use
24243         arc_return_address_register function.
24244         (simple_return): Likewise.
24245         (p_return_i): Likewise.
24247 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24249         PR tree-optimization/81428
24250         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24251         can't be built for those types.
24253 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24255         Remove stuff dead since r239246.
24257         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24258         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24259         (avr_inform_devices): Remove dead stuff.
24261 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24263         * config/arm/arm_neon.h: Fix softp typo.
24265 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24267         PR tree-optimization/81365
24268         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24269         aggregate moves onto bb predecessor edges, make sure there are no
24270         loads that could alias the lhs in between the start of bb and the
24271         loads from *phi.
24273 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24275         PR 80929
24276         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24277         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24278         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24280 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24282         PR tree-optimization/81396
24283         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24284         (init_symbolic_number): Initialize it to 1.
24285         (perform_symbolic_merge): Add n_ops from both operands into the new
24286         n_ops.
24287         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24288         without base_addr as useless if they need more than one operation.
24289         (bswap_replace): Handle !bswap case for NULL base_addr.
24291 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24293         PR target/81069
24294         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24295         as possible.
24297 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24299         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24300         conditional builtin define __FIX_LEON3FT_B2BST.
24302 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24304         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24305         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24306         with -mfix-ut700.
24308 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24310         PR rtl-optimization/81424
24311         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24312         to remove potential trapping from operands if -fnon-call-exceptions.
24314 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24316         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24317         profile_proability for scalling.
24318         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24320 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24322         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24324 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24326         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24327         fixpoint arithmetics.
24329 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24331         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24332         fixpoint arithmetics.
24334 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24336         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24337         fixpoint arithmetics.
24339 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24341         * profile-count.h (profile_probability::from_reg_br_prob_note,
24342         profile_probability::to_reg_br_prob_note): New functions.
24343         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24344         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24345         * predict.c (probability_reliable_p): Update.
24346         (edge_probability_reliable_p): Update.
24347         (br_prob_note_reliable_p): Update.
24348         (invert_br_probabilities): Update.
24349         (add_reg_br_prob_note): New function.
24350         (combine_predictions_for_insn): Update.
24351         * asan.c (asan_clear_shadow): Update.
24352         * cfgbuild.c (compute_outgoing_frequencies): Update.
24353         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24354         (update_br_prob_note): Update.
24355         (rtl_verify_edges): Update.
24356         (purge_dead_edges): Update.
24357         (fixup_reorder_chain): Update.
24358         * emit-rtl.c (try_split): Update.
24359         * ifcvt.c (cond_exec_process_insns): Update.
24360         (cond_exec_process_if_block): Update.
24361         (dead_or_predicable): Update.
24362         * internal-fn.c (expand_addsub_overflow): Update.
24363         (expand_neg_overflow): Update.
24364         (expand_mul_overflow): Update.
24365         * loop-doloop.c (doloop_modify): Update.
24366         * loop-unroll.c (compare_and_jump_seq): Update.
24367         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24368         * predict.h: Update.
24369         * reorg.c (mostly_true_jump): Update.
24370         * rtl.h: Update.
24371         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24372         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24373         * config/arc/arc.c: (emit_unlikely_jump): Update.
24374         * config/arm/arm.c: (emit_unlikely_jump): Update.
24375         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24376         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24377         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24378         (ix86_print_operand): Update.
24379         (ix86_split_fp_branch): Update.
24380         (predict_jump): Update.
24381         * config/ia64/ia64.c (ia64_print_operand): Update.
24382         * config/mmix/mmix.c (mmix_print_operand): Update.
24383         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24384         (rs6000_expand_split_stack_prologue): Update.
24385         * config/rs6000/rs6000.c: Update.
24386         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24387         (s390_expand_vec_movstr): Update.
24388         (s390_expand_cs_tdsi): Update.
24389         (s390_expand_split_stack_prologue): Update.
24390         * config/sh/sh.c (sh_print_operand): Update.
24391         (expand_cbranchsi4): Update.
24392         (expand_cbranchdi4): Update.
24393         * config/sparc/sparc.c (output_v9branch): Update.
24394         * config/spu/spu.c (get_branch_target): Update.
24395         (ea_load_store_inline): Update.
24396         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24397         * config/tilepro/tilepro.c: Update.
24399 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24401         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24402         (gimplify_save_expr): Likewise.
24404 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24406         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24408 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24410         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24411         TV_IPA_FNSUMMARY.
24412         * timevar.def (TV_IPA_FNSUMMARY): Define.
24414 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24416         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24417         to back store errata sensitive sequence from being generated.
24418         (sqrtdf2_fix): Likewise.
24420 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24422         * tree-ssa-threadupdate.c (compute_path_counts,
24423         update_joiner_offpath_counts): Use profile_probability.
24425 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24427         Revert:
24428         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24430         * config/arm/arm-c.c (arm_cpu_builtins): Define
24431         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24433 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24435         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24436         array entries to represent __ieee128 versions of the
24437         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24438         scalar_extract_sig, and scalar_insert_exp built-in functions.
24439         (altivec_resolve_overloaded_builtin): Add special case handling
24440         for the __builtin_scalar_insert_exp function, as represented by
24441         the P9V_BUILTIN_VEC_VSIEDP constant.
24442         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24443         exponent support for __ieee128 argument.
24444         (VSESQP): Add scalar extract signature support for __ieee128
24445         argument.
24446         (VSTDCNQP): Add scalar test negative support for __ieee128
24447         argument.
24448         (VSIEQP): Add scalar insert exponent support for __int128 argument
24449         with __ieee128 result.
24450         (VSIEQPF): Add scalar insert exponent support for __ieee128
24451         argument with __ieee128 result.
24452         (VSTDCQP): Add scalar test data class support for __ieee128
24453         argument.
24454         (VSTDCNQP): Add overload support for scalar test negative with
24455         __ieee128 argument.
24456         (VSTDCQP): Add overload support for scalar test data class
24457         __ieee128 argument.
24458         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24459         UNSPEC_VSX_SXSIGDP.
24460         (UNSPEC_VSX_SIEXPQP): New constant.
24461         (xsxexpqp): New insn for VSX scalar extract exponent quad
24462         precision.
24463         (xsxsigqp): New insn for VSX scalar extract significand quad
24464         precision.
24465         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24466         precision with floating point argument.
24467         (xststdcqp): New expand for VSX scalar test data class quad
24468         precision.
24469         (xststdcnegqp): New expand for VSX scalar test negative quad
24470         precision.
24471         (xststdcqp): New insn to match expansions for VSX scalar test data
24472         class quad precision and VSX scalar test negative quad precision.
24473         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24474         special case operand checking to enforce that second operand of
24475         VSX scalar test data class with quad precision argument is a 7-bit
24476         unsigned literal.
24477         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24478         prototypes and descriptions of __ieee128 versions of
24479         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24480         scalar_test_data_class, and scalar_test_neg built-in functions.
24482 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24484         PR tree-optimization/81162
24485         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24486         replace a negate with an add.
24488 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24490         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24492 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24494         * config/arm/arm-c.c (arm_cpu_builtins): Define
24495         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24497 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24499         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24500         (armv8-r): Set ARM Cortex-R52 as default CPU.
24501         * config/arm/arm-tables.opt: Regenerate.
24502         * config/arm/arm-tune.md: Regenerate.
24503         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24504         Cortex-R52.
24505         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24506         extension for -mcpu=cortex-r52.
24508 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24510         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24511         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24512         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24513         (fp-armv8): Define it as FP_ARMv8 only.
24514         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24515         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24516         TARGET_FPU_ARMV8.
24517         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24518         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24519         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24520         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24521         than TARGET_FPU_ARMV8.
24522         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24523         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24524         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24525         TARGET_FPU_ARMV8.
24526         * config/arm/neon.md (neon_vrint): Likewise.
24527         (neon_vcvt): Likewise.
24528         (neon_<fmaxmin_op><mode>): Likewise.
24529         (<fmaxmin><mode>3): Likewise.
24530         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24531         * config/arm/predicates.md (arm_cond_move_operator): Check against
24532         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24534 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24536         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24537         to top of function.
24539 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24541         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24542         loop in comment with memset.
24544 2017-07-14  Martin Liska  <mliska@suse.cz>
24546         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24547         * dwarf2out.c (is_java): Remove the function.
24548         (output_pubname): Remove usage of the function.
24549         (lower_bound_default): Remove usage of DW_LANG_Java.
24550         (gen_compile_unit_die): Likewise.
24551         * gcc.c: Remove compiler defaults for .java and .zip files.
24552         * gimple-expr.c (remove_suffix): Change as there's no longer
24553         extension than 4-letter one.
24554         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24555         (gimplify_save_expr): Likewise.
24556         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24557         as it's possible even for other languages than Java.
24558         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24559         * lto-opts.c (lto_write_options): Remove reference to Java.
24560         * opts.c (strip_off_ending): Update file extension handling.
24561         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24562         * tree-eh.c (lower_resx): Likewise.
24563         * tree.c (free_lang_data_in_type): Remove dead code.
24564         (find_decls_types_r): Likewise.
24565         (build_common_builtin_nodes): Remove Java from a comment.
24566         (verify_type): Remove dead code.
24567         * varasm.c (assemble_external): Remove Java from a comment.
24569 2017-07-14  Martin Liska  <mliska@suse.cz>
24571         * opts.c (finish_options): Add quotes.
24572         (common_handle_option): Likewise.
24574 2017-07-14  Martin Liska  <mliska@suse.cz>
24576         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24577         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24578         Remove N_SO_PASCAL.
24579         * dwarf2out.c (lower_bound_default): Do not handle
24580         DW_LANG_Pascal83.
24581         (gen_compile_unit_die): Likewise.
24582         * gcc.c: Remove default extension binding for GNU Pascal.
24583         * stmt.c: Remove Pascal language from a comment.
24584         * xcoffout.c: Likewise.
24586 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24588         PR c/81405
24589         * diagnostic-show-locus.c (fixit_cmp): New function.
24590         (layout::layout): Sort m_fixit_hints.
24591         (column_range::column_range): Assert that the values are valid.
24592         (struct char_span): New struct.
24593         (correction::overwrite): New method.
24594         (struct source_line): New struct.
24595         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24596         calls in terms of classes source_line and char_span, and
24597         correction::overwrite.
24598         (selftest::test_overlapped_fixit_printing_2): New function.
24599         (selftest::diagnostic_show_locus_c_tests): Call it.
24601 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24603         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24604         early if there is no lhs.
24606 2017-07-13  Martin Liska  <mliska@suse.cz>
24608         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24609         (gen_reference_type_die): Likewise.
24610         * stor-layout.c: Remove Pascal-related comment.
24612 2017-07-13  Martin Liska  <mliska@suse.cz>
24614         * opts.c (finish_options): Add quotes to error messages.
24615         (parse_sanitizer_options): Likewise.
24617 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24619         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24621 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24623         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24625 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24627         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24628         during expansion.
24629         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24631 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24633         PR target/81193
24634         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24635         provides the hardware capability bits, define the macro
24636         __BUILTIN_CPU_SUPPORTS__.
24637         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24638         if GLIBC does not provide the hardware capability bits.  Add a
24639         gcc_unreachable call if the built-in cpu function is neither
24640         __builtin_cpu_is nor __builtin_cpu_supports.
24641         (rs6000_get_function_versions_dispatcher): Change the warning
24642         that an old GLIBC is used which does not export the capability
24643         bits to be an error.
24644         * doc/extend.texi (target_clones attribute): Document the
24645         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24646         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24647         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24648         the macros defined by GCC if the newer GLIBC is available.
24650 2017-07-12  Jeff Law  <law@redhat.com>
24652         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24653         remaining includes slightly.
24654         * config/riscv/riscv-builtins.c: Include profile-count.h.
24656 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24658         PR target/79883
24659         * config/avr/avr.c (avr_set_current_function): In diagnostic
24660         messages: Quote keywords and (parts of) identifiers.
24661         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24662         "INTERUPT".
24664 2017-07-12  Carl Love  <cel@us.ibm.com>
24666         * config/rs6000/rs6000-c.c: Add support for built-in functions
24667         vector bool char vec_revb (vector bool char);
24668         vector bool short vec_revb (vector short char);
24669         vector bool int vec_revb (vector bool int);
24670         vector bool long long vec_revb (vector bool long long);
24671         * doc/extend.texi: Update the built-in documentation file for the
24672         new built-in functions.
24674 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24676         * config/s390/s390.md: Remove movcc splitter.
24678 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24680         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24681         load/store on condition.
24683 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24685         PR target/81407
24686         * config/avr/avr.c (avr_encode_section_info)
24687         [progmem && !TREE_READONLY]: Error if progmem object needs
24688         constructing.
24690 2017-07-11  Michael Collison  <michael.collison@arm.com>
24692         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24693         New pattern.
24695 2017-07-11  Carl Love  <cel@us.ibm.com>
24697         * config/rs6000/rs6000-c.c: Add support for builtins
24698         vector unsigned int vec_parity_lsbb (vector signed int);
24699         vector unsigned int vec_parity_lsbb (vector unsigned int);
24700         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24701         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24702         vector unsigned long long vec_parity_lsbb (vector signed long long);
24703         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24704         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24705         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24706         * doc/extend.texi: Update the built-in documentation file for the
24707         new built-in functions.
24709 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24711         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24712         (layout::m_primary_loc): New field.
24713         (layout::layout): Initialize new field.  Move location filtering
24714         logic from here to...
24715         (layout::maybe_add_location_range): ...this new method.  Add
24716         support for filtering to just the lines already specified by other
24717         locations.
24718         (layout::will_show_line_p): New method.
24719         (gcc_rich_location::add_location_if_nearby): New method.
24720         (selftest::test_add_location_if_nearby): New test function.
24721         (selftest::diagnostic_show_locus_c_tests): Call it.
24722         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24723         New method.
24725 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24727         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24728         (bb_first_real_insn): New function.
24729         (nvptx_single): Add extra initialization of broadcasted condition
24730         variables.
24732 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24734         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24736 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24738         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24739         Binutils doc as TEXI will mess them up.
24740         * doc/invoke.texi (AVR Options): Same here.
24742 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24744         * config/sparc/sparc.opt (mfix-ut700): New option.
24745         (mfix-gr712rc): Likewise.
24746         (sparc_fix_b2bst): New variable.
24747         * doc/invoke.texi (SPARC options): Document them.
24748         (ARM options): Fix warnings.
24749         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24750         instructions to prevent sequences that can trigger the store-store
24751         errata for certain LEON3FT processors.
24752         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24753         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24754         * config/sparc/sparc.md (fix_b2bst): New attribute.
24755         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24757 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24759         PR target/81375
24760         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24761         (rcpps): Ditto.
24762         (*rsqrtsf2_sse): Ditto.
24763         (rsqrtsf2): Ditto.
24764         (div<mode>3): Macroize insn from divdf3 and divsf3
24765         using MODEF mode iterator.
24767 2017-07-10  Martin Sebor  <msebor@redhat.com>
24769         PR tree-optimization/80397
24770         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24771         instead of testing for equality to INTEGER_TYPE.
24773 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24775         * config.gcc: Remove uclibc from arc target spec.
24777 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24779         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24781 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24783         PR lto/80838
24784         * lto-wrapper.c (remove_option): New function.
24785         (merge_and_complain): Merge PIC/PIE options more realistically.
24787 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24789         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24791         PR target/20296
24792         PR target/81268
24793         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24794         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24795         * config.in: Regenerate.
24796         * configure: Regenerate.
24797         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24798         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24799         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24800         (TARGET_GASISR_PROLOGUES): ...target mask.
24801         * common/config/avr/avr-common.c
24802         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24803         Set -mgas-isr-prologues.
24804         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24805         INSERT_PASS_BEFORE for it.
24806         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24807         * config/avr/avr.c (avr_option_override)
24808         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24809         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24810         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24811         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24812         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24813         and rtl_opt_pass.
24814         (make_avr_pass_pre_proep): New function.
24815         (emit_push_sfr) <treg>: Add argument to function and use it
24816         instead of TMP_REG.
24817         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24818         and set machine->gasisr.yes.
24819         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24820         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24821         __gcc_isr.n_pushed to .L__stack_usage.
24822         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24823         (avr_asm_final_postscan_insn): ...this new static function.
24824         * config/avr/avr.h (machine_function)
24825         <is_no_gccisr, use_L__stack_usage>: New fields.
24826         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24827         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24828         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24829         (gasisr, *gasisr): New expander and insn.
24830         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24831         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24832         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24834 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24836         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24837         in quoted strings.
24839 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24841         Move jump-tables out of .text again.
24843         PR target/81075
24844         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24845         (ASM_OUTPUT_ADDR_VEC): New function.
24846         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24847         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24848         INSN_ADDRESSes as asm comment.
24849         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24850         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24851         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24852         * config/avr/avr.md (*tablejump): Adjust comment.
24853         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24854         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24855         New detail.
24856         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24857         (avr_output_addr_vec): New proto.
24858         (avr_log_t) <insn_addresses>: New field.
24860 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
24862         PR target/81313
24863         * config/i386/i386.c (ix86_function_arg_advance): Set
24864         outgoing_args_on_stack to true if there are outgoing arguments
24865         on stack.
24866         (ix86_function_arg): Likewise.
24867         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
24868         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
24869         * config/i386/i386.h (machine_function): Add
24870         outgoing_args_on_stack.
24872 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
24874         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
24875         supporting pthreds.
24876         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
24878 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
24880         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
24881         (REAL_H): Remove $(MACHMODE_H).
24882         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
24883         double-int.h.
24884         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
24885         $(MACHMODE_H) and double-int.h.
24886         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
24887         $(MACHMODE_H).
24888         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
24889         double-int.h.
24891 2017-07-07  Andrew Pinski  <apinski@cavium.com>
24893         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
24894         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
24896 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24898         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
24899         Add warning if GCC was not configured to link against a GLIBC that
24900         exports the hardware capability bits.
24901         (make_resolver_func): Make resolver function private and not a
24902         COMDAT function.  Create the name with clone_function_name instead
24903         of make_unique_name.
24905         PR target/81348
24906         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
24907         correct operand in doing the split.
24909 2017-07-07 Carl Love  <cel@us.ibm.com>
24911         * config/rs6000/rs6000-c: Add support for built-in function
24912         vector unsigned short vec_pack_to_short_fp32 (vector float,
24913                                                       vector float).
24914         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
24915         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
24916         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
24917         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
24918         (convert_4f32_8i16): Add define_expand.
24919         * doc/extend.texi: Update the built-in documentation file for the
24920         new built-in function.
24922 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24924         * config/sparc/m8.md: New file.
24925         * config/sparc/sparc.md: Include m8.md.
24927 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24929         * config/sparc/sparc.opt: New option -mvis4b.
24930         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
24931         (sparc_option_override): Handle VIS4B.
24932         (enum sparc_builtins): Define
24933         SPARC_BUILTIN_DICTUNPACK{8,16,32},
24934         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
24935         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
24936         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
24937         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
24938         (check_constant_argument): New function.
24939         (sparc_vis_init_builtins): Define builtins
24940         __builtin_vis_dictunpack{8,16,32},
24941         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
24942         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
24943         __builtin_vis_fpcmpde{8,16,32}shl and
24944         __builtin_vis_fpcmpur{8,16,32}shl.
24945         (sparc_expand_builtin): Check that the constant operands to
24946         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
24947         constant and in range.
24948         * config/sparc/sparc-c.c (sparc_target_macros): Handle
24949         TARGET_VIS4B.
24950         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
24951         (SPARC_IMM5_P): Likewise.
24952         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
24953         (enabled): Handle vis4b.
24954         (UNSPEC_DICTUNPACK): New unspec.
24955         (UNSPEC_FPCMPSHL): Likewise.
24956         (UNSPEC_FPUCMPSHL): Likewise.
24957         (UNSPEC_FPCMPDESHL): Likewise.
24958         (UNSPEC_FPCMPURSHL): Likewise.
24959         (cpu_feature): New CPU feature `vis4b'.
24960         (dictunpack{8,16,32}): New insns.
24961         (FPCSMODE): New mode iterator.
24962         (fpcscond): New code iterator.
24963         (fpcsucond): Likewise.
24964         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
24965         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
24966         (fpcmpde{8,16,32}{si,di}shl): Likewise.
24967         (fpcmpur{8,16,32}{si,di}shl): Likewise.
24968         * config/sparc/constraints.md: Define constraints `q' for unsigned
24969         2-bit integer constants and `t' for unsigned 5-bit integer
24970         constants.
24971         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
24972         predicate.
24973         (imm5_operand_dictunpack16): Likewise.
24974         (imm5_operand_dictunpack32): Likewise.
24975         (imm2_operand): Likewise.
24976         * doc/invoke.texi (SPARC Options): Document -mvis4b.
24977         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
24978         ditunpack* and fpcmp*shl builtins.
24980 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24982         * config.gcc: Handle m8 in --with-{cpu,tune} options.
24983         * config.in: Add HAVE_AS_SPARC6 define.
24984         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
24985         M8.
24986         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
24987         TARGET_CPU_m8.
24988         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
24989         (CPP_CPU_SPEC): Handle m8.
24990         (ASM_CPU_SPEC): Likewise.
24991         * config/sparc/sparc-opts.h (enum processor_type): Add
24992         PROCESSOR_M8.
24993         * config/sparc/sparc.c (m8_costs): New struct.
24994         (sparc_option_override): Handle TARGET_CPU_m8.
24995         (sparc32_initialize_trampoline): Likewise.
24996         (sparc64_initialize_trampoline): Likewise.
24997         (sparc_issue_rate): Likewise.
24998         (sparc_register_move_cost): Likewise.
24999         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25000         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25001         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25002         (CPP_CPU_SPEC): Handle M8.
25003         (ASM_CPU_SPEC): Likewise.
25004         (AS_M8_FLAG): Define.
25005         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25006         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25007         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25008         M8 instructions.
25009         * configure: Regenerate.
25010         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25011         -mtune=m8.
25013 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25015         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25016         subtypes.
25017         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25018         ("*movdi_insn_sp32"): Do not set v3pipe.
25019         ("*movsi_insn"): Likewise.
25020         ("*movdi_insn_sp64"): Likewise.
25021         ("*movsf_insn"): Likewise.
25022         ("*movdf_insn_sp32"): Likewise.
25023         ("*movdf_insn_sp64"): Likewise.
25024         ("*zero_extendsidi2_insn_sp64"): Likewise.
25025         ("*sign_extendsidi2_insn"): Likewise.
25026         ("*mov<VM32:mode>_insn"): Likewise.
25027         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25028         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25029         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25030         ("<vlop:code><VL:mode>3"): Likewise.
25031         ("*not_<vlop:code><VL:mode>3"): Likewise.
25032         ("*nand<VL:mode>_vis"): Likewise.
25033         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25034         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25035         ("one_cmpl<VL:mode>2"): Likewise.
25036         ("faligndata<VM64:mode>_vis"): Likewise.
25037         ("alignaddrsi_vis"): Likewise.
25038         ("alignaddrdi_vis"): Likweise.
25039         ("alignaddrlsi_vis"): Likewise.
25040         ("alignaddrldi_vis"): Likewise.
25041         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25042         ("bmaskdi_vis"): Likewise.
25043         ("bmasksi_vis"): Likewise.
25044         ("bshuffle<VM64:mode>_vis"): Likewise.
25045         ("cmask8<P:mode>_vis"): Likewise.
25046         ("cmask16<P:mode>_vis"): Likewise.
25047         ("cmask32<P:mode>_vis"): Likewise.
25048         ("pdistn<P:mode>_vis"): Likewise.
25049         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25051 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25053         * config/sparc/sparc.md ("subtype"): New insn attribute.
25054         ("*wrgsr_sp64"): Set insn subtype.
25055         ("*rdgsr_sp64"): Likewise.
25056         ("alignaddrsi_vis"): Likewise.
25057         ("alignaddrdi_vis"): Likewise.
25058         ("alignaddrlsi_vis"): Likewise.
25059         ("alignaddrldi_vis"): Likewise.
25060         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25061         ("fexpand_vis"): Likewise.
25062         ("fpmerge_vis"): Likewise.
25063         ("faligndata<VM64:mode>_vis"): Likewise.
25064         ("bshuffle<VM64:mode>_vis"): Likewise.
25065         ("cmask8<P:mode>_vis"): Likewise.
25066         ("cmask16<P:mode>_vis"): Likewise.
25067         ("cmask32<P:mode>_vis"): Likewise.
25068         ("fchksm16_vis"): Likewise.
25069         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25070         ("fmean16_vis"): Likewise.
25071         ("fp<plusminus_insn>64_vis"): Likewise.
25072         ("<plusminus_insn>v8qi3"): Likewise.
25073         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25074         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25075         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25076         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25077         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25078         ("*movqi_insn"): Likewise.
25079         ("*movhi_insn"): Likewise.
25080         ("*movsi_insn"): Likewise.
25081         ("movsi_pic_gotdata_op"): Likewise.
25082         ("*movdi_insn_sp32"): Likewise.
25083         ("*movdi_insn_sp64"): Likewise.
25084         ("movdi_pic_gotdata_op"): Likewise.
25085         ("*movsf_insn"): Likewise.
25086         ("*movdf_insn_sp32"): Likewise.
25087         ("*movdf_insn_sp64"): Likewise.
25088         ("*zero_extendhisi2_insn"): Likewise.
25089         ("*zero_extendqihi2_insn"): Likewise.
25090         ("*zero_extendqisi2_insn"): Likewise.
25091         ("*zero_extendqidi2_insn"): Likewise.
25092         ("*zero_extendhidi2_insn"): Likewise.
25093         ("*zero_extendsidi2_insn_sp64"): Likewise.
25094         ("ldfsr"): Likewise.
25095         ("prefetch_64"): Likewise.
25096         ("prefetch_32"): Likewise.
25097         ("tie_ld32"): Likewise.
25098         ("tie_ld64"): Likewise.
25099         ("*tldo_ldub_sp32"): Likewise.
25100         ("*tldo_ldub1_sp32"): Likewise.
25101         ("*tldo_ldub2_sp32"): Likewise.
25102         ("*tldo_ldub_sp64"): Likewise.
25103         ("*tldo_ldub1_sp64"): Likewise.
25104         ("*tldo_ldub2_sp64"): Likewise.
25105         ("*tldo_ldub3_sp64"): Likewise.
25106         ("*tldo_lduh_sp32"): Likewise.
25107         ("*tldo_lduh1_sp32"): Likewise.
25108         ("*tldo_lduh_sp64"): Likewise.
25109         ("*tldo_lduh1_sp64"): Likewise.
25110         ("*tldo_lduh2_sp64"): Likewise.
25111         ("*tldo_lduw_sp32"): Likewise.
25112         ("*tldo_lduw_sp64"): Likewise.
25113         ("*tldo_lduw1_sp64"): Likewise.
25114         ("*tldo_ldx_sp64"): Likewise.
25115         ("*mov<VM32:mode>_insn"): Likewise.
25116         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25117         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25119 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25121         * config/sparc/sparc.md ("type"): New insn type viscmp.
25122         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25123         viscmp.
25124         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25125         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25126         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25127         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25128         viscmp.
25129         ("n7_vis_logical_11cycle"): Likewise.
25130         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25131         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25132         * config/sparc/niagara.md ("niag_vis"): Likewise.
25133         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25134         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25136 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25138         * config/sparc/sparc.md: New instruction type `bmask'.
25139         (bmaskdi_vis): Use the `bmask' type.
25140         (bmasksi_vis): Likewise.
25141         * config/sparc/ultra3.md (us3_array): Likewise.
25142         * config/sparc/niagara7.md (n7_array): Likewise.
25143         * config/sparc/niagara4.md (n4_array): Likewise.
25144         * config/sparc/niagara2.md (niag2_vis): Likewise.
25145         (niag3_vis): Likewise.
25146         * config/sparc/niagara.md (niag_vis): Likewise.
25148 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25150         * ipa-comdats.c: Remove optimize check from gate.
25151         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25152         for functions not optimized.
25153         (ipa_fn_summary_read): Skip optimize check.
25154         (ipa_fn_summary_write): Likewise.
25155         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25156         is optimized.
25157         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25158         uninlinable.
25159         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25160         (check_callers): Give up on caller which is not optimized.
25161         (inline_small_functions): Likewise.
25162         (ipa_inline): Do not give up when not optimizing.
25163         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25164         away unoptimizes cdtors.
25165         (whole_program_function_and_variable_visibility): Do
25166         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25167         * ipa.c (process_references): Do not check optimize.
25168         (symbol_table::remove_unreachable_nodes): Update optimize check.
25169         (set_writeonly_bit): Update optimize check.
25170         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25171         (pass_ipa_single_use::gate): Remove.
25173 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25175         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25176         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25177         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25178         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25179         permute_load, permute_store, adjust_extract, adjust_splat,
25180         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25181         replace_swap_with_copy, dump_swap_insn_table,
25182         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25183         recombine_lvx_pattern, recombine_stvx_pattern,
25184         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25185         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25186         to file rs6000-p8swap.c.
25187         * config/rs6000/rs6000-p8swap.c: New file.
25188         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25189         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25190         and rs6000*-*-* targets.
25192 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25194         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25196 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25198         * lto-wrapper.c (merge_and_complain): Do not merge
25199         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25200         fsigned_zeros, ftrapping_math, fwrapv.
25201         (append_compiler_options): Do not track these options.
25202         (append_linker_options): Likewie
25204 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25206         * cgraphunit.c (cgraph_node::finalize_function): When
25207         !flag_toplevel_reorde set no_reorder flag.
25208         (varpool_node::finalize_decl): Likewise.
25209         (symbol_table::compile): Drop no toplevel reorder path.
25211 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25213         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25214         edges; zero probability is not better than uninitialized.
25216 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25218         * asan.h (asan_sanitize_allocas_p): Declare.
25219         * asan.c (asan_sanitize_allocas_p): New function.
25220         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25221         (handle_builtin_alloca): Likewise.
25222         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25223         if !asan_sanitize_allocas_p.
25224         * params.def (asan-instrument-allocas): Add new option.
25225         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25226         * opts.c (common_handle_option): Disable allocas sanitization for
25227         KASan by default.
25229 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25231         * asan.c: Include gimple-fold.h.
25232         (get_last_alloca_addr): New function.
25233         (handle_builtin_stackrestore): Likewise.
25234         (handle_builtin_alloca): Likewise.
25235         (asan_emit_allocas_unpoison): Likewise.
25236         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25237         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25238         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25239         (instrument_builtin_call): Pass gimple iterator to
25240         get_mem_refs_of_builtin_call.
25241         (last_alloca_addr): New global.
25242         * asan.h (asan_emit_allocas_unpoison): Declare.
25243         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25244         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25245         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25246         if function calls alloca.
25247         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25248         * gimple-fold.h (replace_call_with_value): Declare.
25249         * internal-fn.c: Include asan.h.
25250         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25251         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25253 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25255         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25256         (C_SELFTEST_FLAGS): New.
25257         (CPP_SELFTEST_FLAGS): New.
25258         (SELFTEST_DEPS): New, from deps of s-selftest.
25259         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25260         (CPP_SELFTEST_DEPS): New.
25261         (selftest): Add dependency on s-selftest-c++.
25262         (s-selftest): Rename to...
25263         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25264         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25265         than SELFTEST_FLAGS.
25266         (selftest-gdb): Rename to...
25267         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25268         C_SELFTEST_FLAGS.
25269         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25270         (selftest-valgrind): Rename to...
25271         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25272         C_SELFTEST_FLAGS.
25273         (selftest-valgrind): Reintroduce as an alias for
25274         selftest-c-valgrind.
25275         (s-selftest-c++): New.
25276         (selftest-c++-gdb): New.
25277         (selftest-c++-valgrind): New.
25279 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25281         * gcc.c (process_command): When deciding if undefined variables
25282         should be ignored when processing specs, accept "gcc -v" as well.
25284 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25286         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25287         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25289 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25291         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25292         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25293         * config/arm/arm-tables.opt: Regenerate.
25294         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25295         enumerator.
25296         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25298 2017-07-06  Carl Love  <cel@us.ibm.com>
25300         * ChangeLog: Clean up from mid air collision
25302 2017-07-06  Carl Love  <cel@us.ibm.com>
25304         * config/rs6000/rs6000-c.c: Add support for built-in functions
25305         vector signed int vec_subc (vector signed int, vector signed int);
25306         vector signed __int128 vec_subc (vector signed __int128,
25307                                          vector signed __int128);
25308         vector unsigned __int128 vec_subc (vector unsigned __int128,
25309                                            vector unsigned __int128);
25310         vector signed int vec_sube (vector signed int, vector signed int,
25311                                     vector signed int);
25312         vector unsigned int vec_sube (vector unsigned int,
25313                                       vector unsigned int,
25314                                       vector unsigned int);
25315         vector signed __int128 vec_sube (vector signed __int128,
25316                                          vector signed __int128,
25317                                          vector signed__int128);
25318         vector unsigned __int128 vec_sube (vector unsigned __int128,
25319                                            vector unsigned __int128,
25320                                            vector unsigned __int128);
25321         vector signed int vec_subec (vector signed int, vector signed int,
25322                                      vector signed int);
25323         vector unsigned int vec_subec (vector unsigned int,
25324                                        vector unsigned int,
25325                                        vector unsigned int);
25326         vector signed __int128 vec_subec (vector signed __int128,
25327                                           vector signed __int128,
25328                                           vector signed__int128);
25329         vector unsigned __int128 vec_subec (vector unsigned __int128,
25330                                             vector unsigned __int128,
25331                                             vector unsigned __int128);
25332         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25333         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25334         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25335         BU_ALTIVEC_OVERLOAD_X definitions.
25336         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25337         * doc/extend.texi: Update the built-in documentation file for the new
25338         built-in functions.
25340 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25342         PR c++/79300
25343         * diagnostic-show-locus.c (layout::layout): Use start and finish
25344         spelling location for the start and finish of each range.
25345         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25346         Add unused aspect param.
25347         * input.c (expand_location_1): Add "aspect" param, and use it
25348         to access the correct part of the location.
25349         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25350         expand_location_1.
25351         (expand_location_to_spelling_point): Likewise.
25352         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25353         param, and pass it to expand_location_1.
25355 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25357         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25358         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25359         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25360         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25361         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25362         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25363         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25364         _mm_maskz_getmant_ss): New intrinsics.
25365         (__builtin_ia32_getexpss128_mask): Changed to ...
25366         __builtin_ia32_getexpss128_round ... this.
25367         (__builtin_ia32_getexpsd128_mask): Changed to ...
25368         __builtin_ia32_getexpsd128_round ... this.
25369         * config/i386/i386-builtin-types.def
25370         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25371         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25372         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25373         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25374         __builtin_ia32_getmantss_mask_round): New builtins.
25375         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25376         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25377         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25378         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25379         * config/i386/sse.md
25380         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25381         avx512f_sgetexp<mode><mask_scalar_name>
25382         <round_saeonly_scalar_name> ... this.
25383         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25384         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25385         vgetexp<ssescalarmodesuffix>
25386         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25387         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25388         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25389         avx512f_vgetmant<mode><mask_scalar_name>
25390         <round_saeonly_scalar_name> ... this.
25391         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25392         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25393         vgetmant<ssescalarmodesuffix>
25394         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25395         %0<mask_scalar_operand4>, %1, %2
25396         <round_saeonly_scalar_mask_op4>, %3} ... this.
25397         * config/i386/subst.md (mask_scalar_operand4,
25398         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25399         round_saeonly_scalar_nimm_predicate): New subst attributes.
25401 2017-07-06  Julia Koval  <julia.koval@intel.com>
25403         * config/i386/i386.c (ix86_erase_embedded_rounding):
25404         Remove code for old rounding pattern.
25406 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25408         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25410 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25412         * doc/sourcebuild.texi (Test Directives, Variants of
25413         dg-require-support): Add documentation for dg-require-stack-check.
25415 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25417         * config/i386/subst.md (mask_scalar, round_scalar,
25418         round_saeonly_scalar): New meta-templates.
25419         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25420         round_scalar_mask_operand3, round_scalar_mask_op3,
25421         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25422         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25423         round_saeonly_scalar_constraint,
25424         round_saeonly_scalar_prefix): New subst attribute.
25425         * config/i386/sse.md
25426         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25427         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25428         <round_scalar_name> ... this.
25429         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25430         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25431         <round_scalar_name> ... this.
25432         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25433         <sse>_vm<code><mode>3<mask_scalar_name>
25434         <round_saeonly_scalar_name> ... this.
25435         (v<plusminus_mnemonic><ssescalarmodesuffix>
25436         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25437         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25438         v<plusminus_mnemonic><ssescalarmodesuffix>
25439         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25440         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25441         (v<multdiv_mnemonic><ssescalarmodesuffix>
25442         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25443         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25444         v<multdiv_mnemonic><ssescalarmodesuffix>
25445         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25446         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25447         (v<maxmin_float><ssescalarmodesuffix>
25448         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25449         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25450         v<maxmin_float><ssescalarmodesuffix>
25451         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25452         %0<mask_scalar_operand3>, %1, %<iptr>2
25453         <round_saeonly_scalar_mask_op3>} ... this.
25455 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25457         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25458         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25460 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25461             Alan Hayward  <alan.hayward@arm.com>
25462             David Sherwood  <david.sherwood@arm.com>
25464         * combine.c (simplify_if_then_else): Remove "enum" before
25465         "machine_mode".
25466         * compare-elim.c (can_eliminate_compare): Likewise.
25467         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25468         Likewise.
25469         (aarch64_lookup_simd_builtin_type): Likewise.
25470         (aarch64_simd_builtin_type): Likewise.
25471         (aarch64_init_simd_builtin_types): Likewise.
25472         (aarch64_simd_expand_args): Likewise.
25473         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25474         Likewise.
25475         (aarch64_reverse_mask): Likewise.
25476         (aarch64_simd_emit_reg_reg_move): Likewise.
25477         (aarch64_gen_adjusted_ldpstp): Likewise.
25478         (aarch64_ccmp_mode_to_code): Likewise.
25479         (aarch64_operands_ok_for_ldpstp): Likewise.
25480         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25481         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25482         Likewise.
25483         (aarch64_min_divisions_for_recip_mul): Likewise.
25484         (aarch64_reassociation_width): Likewise.
25485         (aarch64_get_condition_code_1): Likewise.
25486         (aarch64_simd_emit_reg_reg_move): Likewise.
25487         (aarch64_simd_attr_length_rglist): Likewise.
25488         (aarch64_reverse_mask): Likewise.
25489         (aarch64_operands_ok_for_ldpstp): Likewise.
25490         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25491         (aarch64_gen_adjusted_ldpstp): Likewise.
25492         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25493         Likewise.
25494         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25495         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25496         (arm_lookup_simd_builtin_type): Likewise.
25497         (arm_simd_builtin_type): Likewise.
25498         (arm_init_simd_builtin_types): Likewise.
25499         (arm_expand_builtin_args): Likewise.
25500         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25501         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25502         (ft32_setup_incoming_varargs): Likewise.
25503         (ft32_function_arg): Likewise.
25504         (ft32_function_arg_advance): Likewise.
25505         (ft32_pass_by_reference): Likewise.
25506         (ft32_arg_partial_bytes): Likewise.
25507         (ft32_valid_pointer_mode): Likewise.
25508         (ft32_addr_space_pointer_mode): Likewise.
25509         (ft32_addr_space_legitimate_address_p): Likewise.
25510         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25511         Likewise.
25512         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25513         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25514         (iamcu_alignment): Likewise.
25515         (canonicalize_vector_int_perm): Likewise.
25516         (ix86_noce_conversion_profitable_p): Likewise.
25517         (ix86_mpx_bound_mode): Likewise.
25518         (ix86_operands_ok_for_move_multiple): Likewise.
25519         * config/microblaze/microblaze-protos.h
25520         (microblaze_expand_conditional_branch_reg): Likewise.
25521         * config/microblaze/microblaze.c
25522         (microblaze_expand_conditional_branch_reg): Likewise.
25523         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25524         Likewise.
25525         (rs6000_reassociation_width): Likewise.
25526         (rs6000_invalid_binary_op): Likewise.
25527         (fusion_p9_p): Likewise.
25528         (emit_fusion_p9_load): Likewise.
25529         (emit_fusion_p9_store): Likewise.
25530         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25531         Likewise.
25532         (riscv_hard_regno_mode_ok_p): Likewise.
25533         (riscv_address_insns): Likewise.
25534         (riscv_split_symbol): Likewise.
25535         (riscv_legitimize_move): Likewise.
25536         (riscv_function_value): Likewise.
25537         (riscv_hard_regno_nregs): Likewise.
25538         (riscv_expand_builtin): Likewise.
25539         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25540         (riscv_build_integer): Likewise.
25541         (riscv_split_integer): Likewise.
25542         (riscv_legitimate_constant_p): Likewise.
25543         (riscv_cannot_force_const_mem): Likewise.
25544         (riscv_regno_mode_ok_for_base_p): Likewise.
25545         (riscv_valid_base_register_p): Likewise.
25546         (riscv_valid_offset_p): Likewise.
25547         (riscv_valid_lo_sum_p): Likewise.
25548         (riscv_classify_address): Likewise.
25549         (riscv_legitimate_address_p): Likewise.
25550         (riscv_address_insns): Likewise.
25551         (riscv_load_store_insns): Likewise.
25552         (riscv_force_binary): Likewise.
25553         (riscv_split_symbol): Likewise.
25554         (riscv_force_address): Likewise.
25555         (riscv_legitimize_address): Likewise.
25556         (riscv_move_integer): Likewise.
25557         (riscv_legitimize_const_move): Likewise.
25558         (riscv_legitimize_move): Likewise.
25559         (riscv_address_cost): Likewise.
25560         (riscv_subword): Likewise.
25561         (riscv_output_move): Likewise.
25562         (riscv_canonicalize_int_order_test): Likewise.
25563         (riscv_emit_int_order_test): Likewise.
25564         (riscv_function_arg_boundary): Likewise.
25565         (riscv_pass_mode_in_fpr_p): Likewise.
25566         (riscv_pass_fpr_single): Likewise.
25567         (riscv_pass_fpr_pair): Likewise.
25568         (riscv_get_arg_info): Likewise.
25569         (riscv_function_arg): Likewise.
25570         (riscv_function_arg_advance): Likewise.
25571         (riscv_arg_partial_bytes): Likewise.
25572         (riscv_function_value): Likewise.
25573         (riscv_pass_by_reference): Likewise.
25574         (riscv_setup_incoming_varargs): Likewise.
25575         (riscv_print_operand): Likewise.
25576         (riscv_elf_select_rtx_section): Likewise.
25577         (riscv_save_restore_reg): Likewise.
25578         (riscv_for_each_saved_reg): Likewise.
25579         (riscv_register_move_cost): Likewise.
25580         (riscv_hard_regno_mode_ok_p): Likewise.
25581         (riscv_hard_regno_nregs): Likewise.
25582         (riscv_class_max_nregs): Likewise.
25583         (riscv_memory_move_cost): Likewise.
25584         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25585         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25586         (rl78_addr_space_address_mode): Likewise.
25587         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25588         Likewise.
25589         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25590         (rs6000_reassociation_width): Likewise.
25591         (rs6000_invalid_binary_op): Likewise.
25592         (fusion_p9_p): Likewise.
25593         (emit_fusion_p9_load): Likewise.
25594         (emit_fusion_p9_store): Likewise.
25595         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25596         (ok_for_simple_move_operands): Likewise.
25597         (ok_for_simple_move_strict_operands): Likewise.
25598         (ok_for_simple_arith_logic_operands): Likewise.
25599         (visium_legitimize_reload_address): Likewise.
25600         (visium_select_cc_mode): Likewise.
25601         (output_cbranch): Likewise.
25602         (visium_split_double_move): Likewise.
25603         (visium_expand_copysign): Likewise.
25604         (visium_expand_int_cstore): Likewise.
25605         (visium_expand_fp_cstore): Likewise.
25606         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25607         (visium_function_arg): Likewise.
25608         (visium_function_arg_advance): Likewise.
25609         (visium_libcall_value): Likewise.
25610         (visium_setup_incoming_varargs): Likewise.
25611         (visium_legitimate_constant_p): Likewise.
25612         (visium_legitimate_address_p): Likewise.
25613         (visium_legitimize_address): Likewise.
25614         (visium_secondary_reload): Likewise.
25615         (visium_register_move_cost): Likewise.
25616         (visium_memory_move_cost): Likewise.
25617         (prepare_move_operands): Likewise.
25618         (ok_for_simple_move_operands): Likewise.
25619         (ok_for_simple_move_strict_operands): Likewise.
25620         (ok_for_simple_arith_logic_operands): Likewise.
25621         (visium_function_value_1): Likewise.
25622         (rtx_ok_for_offset_p): Likewise.
25623         (visium_legitimize_reload_address): Likewise.
25624         (visium_split_double_move): Likewise.
25625         (visium_expand_copysign): Likewise.
25626         (visium_expand_int_cstore): Likewise.
25627         (visium_expand_fp_cstore): Likewise.
25628         (visium_split_cstore): Likewise.
25629         (visium_select_cc_mode): Likewise.
25630         (visium_split_cbranch): Likewise.
25631         (output_cbranch): Likewise.
25632         (visium_print_operand_address): Likewise.
25633         * expmed.c (flip_storage_order): Likewise.
25634         * expmed.h (emit_cstore): Likewise.
25635         (flip_storage_order): Likewise.
25636         * genrecog.c (validate_pattern): Likewise.
25637         * hsa-gen.c (gen_hsa_addr): Likewise.
25638         * internal-fn.c (expand_arith_overflow): Likewise.
25639         * ira-color.c (allocno_copy_cost_saving): Likewise.
25640         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25641         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25642         (process_invariant_for_inheritance): Likewise.
25643         * lra-eliminations.c (move_plus_up): Likewise.
25644         * omp-low.c (lower_oacc_reductions): Likewise.
25645         * simplify-rtx.c (simplify_subreg): Likewise.
25646         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25647         (TARGET_CHKP_BOUND_MODE): Likewise..
25648         * targhooks.c (default_chkp_bound_mode): Likewise.
25649         (default_setup_incoming_vararg_bounds): Likewise.
25650         * targhooks.h (default_chkp_bound_mode): Likewise.
25651         (default_setup_incoming_vararg_bounds): Likewise.
25652         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25653         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25654         (have_whole_vector_shift): Likewise.
25655         * tree-vect-stmts.c (vectorizable_load): Likewise.
25656         * doc/tm.texi: Regenerate.
25658 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25660         Graceful degrade if Binutils PR21472 is not available.
25662         PR target/81072
25663         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25664         .rodata in flash test fails.
25665         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25666         * confgure: Regenerate.
25667         * config.in: Regenerate.
25668         * config/avr/avr.c (avr_asm_named_section)
25669         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25670         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25671         (avr_asm_init_sections): Same.
25673 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25675         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25676         (fma<VH:mode>4_intrinsic): Likewise.
25677         (*fmsub<VCVTF:mode>4): Likewise.
25678         (*fmsub<VH:mode>4_intrinsic): Likewise.
25680 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25682         PR target/81305
25683         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25684         Don't depend on "optimize > 0".
25685         (out_movhi_r_mr, out_movqi_mr_r): Same.
25686         (out_movhi_mr_r, out_movqi_r_mr): Same.
25687         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25688         io_address_operand on "optimize > 0".
25690 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25692         * tree-loop-distribution.c: Add general explanantion on the pass.
25693         (generate_loops_for_partition): Mark distributed loop.
25694         (pg_add_dependence_edges): New parameter.  Handle alias data
25695         dependence specially and record it in the parameter if asked.
25696         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25697         (init_partition_graph_vertices, add_partition_graph_edge): New.
25698         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25699         (free_partition_graph_vdata, build_partition_graph): New.
25700         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25701         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25702         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25703         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25704         (version_for_distribution_p, finalize_partitions): New.
25705         (distribute_loop): Handle alias data dependence specially.  Factor
25706         out loop fusion code as functions and call these functions.
25708 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25710         * tree-loop-distribution.c (classify_partition): New parameter and
25711         better handle reduction statement.
25712         (rdg_build_partitions): Revise comment.
25713         (distribute_loop): Compute statements in all partitions and pass it
25714         to classify_partition.
25716 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25718         * tree-loop-distribution.c (enum partition_type): New.
25719         (struct partition): New field type.
25720         (partition_merge_into): Add parameter.  Update partition type.
25721         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25722         (build_rdg_partition_for_vertex): Compute partition type.
25723         (rdg_build_partitions): Dump partition type.
25724         (distribute_loop): Update calls to partition_merge_into.
25726 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25728         * tree-loop-distribution.c (struct ddr_hasher): New.
25729         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25730         (ddrs_table): New.
25731         (classify_partition): Call get_data_dependence.
25732         (pg_add_dependence_edges): Ditto.
25733         (distribute_loop): Release data dependence hash table.
25735 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25737         * tree-loop-distribution.c (ref_base_address): Delete.
25738         (similar_memory_accesses): Rename ...
25739         (share_memory_accesses): ... to this.  Check if partitions access
25740         the same memory reference.
25741         (distribute_loop): Call share_memory_accesses.
25743 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25745         * tree-loop-distribution.c (struct partition): New field recording
25746         its data reference.
25747         (partition_alloc, partition_free): Init and release data refs.
25748         (partition_merge_into): Merge data refs.
25749         (build_rdg_partition_for_vertex): Collect data refs for partition.
25750         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25751         Update uses.
25752         (distribute_loop): Remve data refs from vertice data of partition
25753         graph.
25755 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25757         * tree-loop-distribution.c (params.h): Include header file.
25758         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25759         (datarefs_vec): New global var.
25760         (create_rdg_vertices): Use datarefs_vec directly.
25761         (free_rdg): Don't free data references.
25762         (build_rdg): Update use.  Don't free data references.
25763         (distribute_loop): Compute global variable for data references.
25764         Bail out if there are too many data references.
25766 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25768         * tree-loop-distribution.c (loop_nest): New global var.
25769         (build_rdg): Use loop directly, rather than loop nest.
25770         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25771         variable directly.
25772         (distribute_loop): Compute global variable loop nest.  Update use.
25774 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25776         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25777         (partition_merge_into): New parameter.  Dump reason for fusion.
25778         (distribute_loop): Update use of partition_merge_into.
25780 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25782         * tree-loop-distribution.c (bb_top_order_index): New.
25783         (bb_top_order_index_size, bb_top_order_cmp): New.
25784         (stmts_from_loop): Use topological order.
25785         (pass_loop_distribution::execute): Compute and release topological
25786         order for basic blocks.
25788 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25790         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25791         if no loops.
25793 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25795         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25796         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25797         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25798         * internal-fn.def (LOOP_DIST_ALIAS): New.
25799         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25800         (fold_loop_internal_call): ... this.
25801         (vect_loop_dist_alias_call): New function.
25802         (set_uid_loop_bbs): Call fold_loop_internal_call.
25803         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25804         internal calls.
25806 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25808         PR target/81300
25809         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25810         Require dead FLAGS_REG at the beginning of a peephole.
25812 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25814         PR target/81294
25815         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25816         arguments in the call to __builtin_ia32_sbb_u32.
25817         (_subborrow_u64): Swap _X and _Y arguments in the call to
25818         __builtin_ia32_sbb_u64.
25820 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25822         PR debug/81278
25823         * tree-vrp.c (compare_assert_loc): Turn into a function template
25824         with stable template parameter.  Only test if a->e is NULL,
25825         !a->e == !b->e has been verified already.  Use e == NULL or
25826         e != NULL instead of e or ! e tests.  If stable is true, don't use
25827         iterative_hash_expr, on the other side allow a or b or both NULL
25828         and sort the NULLs last.
25829         (process_assert_insertions): Sort using compare_assert_loc<false>
25830         instead of compare_assert_loc, later sort using
25831         compare_assert_loc<true> before calling process_assert_insertions_for
25832         in a loop.  Use break instead of continue once seen NULL pointer.
25834 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25836         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25837         Cortex-R7 and Cortex-R8 processors.
25839 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25841         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25842         uninitialized while src is not.
25844 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25846         * common/config/arm/arm-common.c: Adjust include path for
25847         arm-cpu-cdata.h
25848         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25849         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
25850         (arm-cpu-data.h): Likewise.
25851         (arm-cpu-cdata.h): Likewise.
25852         * config/arm/arm-cpu.h: Delete.
25853         * config/arm/arm-cpu-cdata.h: Delete.
25854         * config/arm/arm-cpu-data.h: Delete.
25856 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
25858         * config/arm/arm-cpus.in (cortex-a55): New.
25859         (cortex-a75): Likewise.
25860         (cortex-a75.cortex-a55): Likewise.
25861         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25862         cortex-a75.
25863         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
25864         * config/arm/arm-cpu-cdata.h: Regenerate.
25865         * config/arm/arm-cpu-data.h: Regenerate.
25866         * config/arm/arm-cpu.h: Regenerate.
25867         * config/arm/arm-tables.opt: Regenerate.
25868         * config/arm/arm-tune.md: Regenerate.
25870 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25872         * haifa-sched.c (sched_create_recovery_edges): Update profile.
25874 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25876         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
25877         probability.
25879 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
25881         PR tree-optimization/81292
25882         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
25883         full_string_p, also call adjust_related_strinfos if the adjustment
25884         is simple, otherwise invalidate related strinfos.
25886 2017-07-04  Martin Liska  <mliska@suse.cz>
25888         PR sanitizer/81040
25889         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
25890         newly created variable as DECL_IGNORED_P.
25892 2017-07-04  Martin Liska  <mliska@suse.cz>
25894         PR ipa/81293
25895         * ipa-inline.c (inline_small_functions):
25896         Use xstrdup_for_dump.
25898 2017-07-04  Tom de Vries  <tom@codesourcery.com>
25900         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
25902 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25904         PR target/81033
25905         * config/darwin.c (darwin_function_switched_text_sections):
25906         Fix spaces.
25908 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
25910         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
25912 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
25914         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
25916 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25918         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
25919         min_profitable_iters, and th as inclusive lower bounds.
25920         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
25921         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
25922         for min_profitable_iters and min_profitable_estimate.
25923         (vect_transform_loop): Treat th as an inclusive lower bound.
25924         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
25926 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25928         PR target/81033
25929         * config/darwin.c (darwin_function_switched_text_sections):
25930         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
25931         in two pieces, and suppress the use of buf.
25933 2017-07-03  Nathan Sidwell  <nathan@acm.org>
25935         * hash-table.h (hash_table_mod1): Fix indentation.
25937 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
25939         PR middle-end/81290
25940         * predict.c (force_edge_cold): Be more careful about propagation
25941         backward.
25942         * profile-count.h (profile_probability::guessed,
25943         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
25944         New.
25945         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
25947 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
25949         * doc/invoke.texi (rcpc architecture extension): Document it.
25951 2017-07-03  Richard Biener  <rguenther@suse.de>
25953         PR tree-optimization/60510
25954         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
25955         the scalar reduction PHI and use it.
25956         (vectorizable_reduction): Properly guard the single_defuse_cycle
25957         path for non-SLP reduction chains where we cannot use it.
25958         Rework reduc_def/index and vector type deduction.  Rework
25959         vector operand gathering during reduction op code-gen.
25960         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
25961         chains dissolve the chain and leave it to non-SLP reduction
25962         handling.
25964 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25966         * tree-data-ref.h (dr_alignment): Declare.
25967         * tree-data-ref.c (dr_alignment): New function.
25968         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
25969         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
25970         set it.
25971         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
25973 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25975         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
25976         and base_misalignment fields.
25977         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
25978         * tree-data-ref.c: Include builtins.h.
25979         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
25980         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
25981         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
25982         * tree-vect-data-refs.c: Include tree-cfg.h.
25983         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
25984         fields instead of calculating an alignment here.
25985         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
25986         innermost_loop_behavior fields.
25988 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25990         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
25991         field.
25992         (DR_STEP_ALIGNMENT): New macro.
25993         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
25994         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
25995         (create_data_ref): Print it.
25996         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
25997         to tell whether the step preserves vector (mis)alignment.
25998         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
25999         Move the check for an integer step and generalise to all INTEGER_CST.
26000         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26001         Print the outer step alignment.
26003 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26005         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26006         with offset_alignment.
26007         (DR_ALIGNED_TO): Delete.
26008         (DR_OFFSET_ALIGNMENT): New macro.
26009         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26010         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26011         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26012         (create_data_ref): Likewise.
26013         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26014         (vect_analyze_data_refs): Likewise.
26015         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26016         creating dummy innermost behavior.
26018 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26020         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26021         with a "innermost_loop_behavior *" and refeence tree.
26022         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26023         (create_data_ref): Update call accordingly.
26024         * tree-predcom.c (find_looparound_phi): Likewise.
26026 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26028         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26029         fields with dr_wrt_vec_loop.
26030         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26031         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26032         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26033         (vect_dr_behavior): New function.
26034         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26035         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26036         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26037         track whether the step preserves the misalignment.
26038         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26039         Use vect_dr_behavior.
26040         (vect_setup_realignment): Update call accordingly.
26041         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26042         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26043         call to vect_create_addr_base_for_vector_ref.
26044         (vect_create_cond_for_align_checks): Likewise.
26045         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26046         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26047         (vect_recog_mask_conversion_pattern): Likewise.
26048         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26049         (new_stmt_vec_info): Remove redundant zeroing.
26051 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26053         * common/config/arm/arm-common.c (arm_be8_option): New function.
26054         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26055         (ISA_ARMv6): Add isa_bit_be8.
26056         * config/arm/arm.h (arm_be8_option): Add prototype.
26057         (BE8_SPEC_FUNCTION): New define.
26058         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26059         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26060         (mlittle-endian): Similarly.
26061         (mbe8, mbe32): New options.
26062         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26063         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26065 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26067         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26069 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26071         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26072         (cleanup_tree_cfg_bb): Use it.
26073         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26074         New functions.
26075         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26077 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26079         PR bootstrap/81285
26080         * loop-doloop.c (add_test): Update profile.
26082 2017-07-03  Martin Liska  <mliska@suse.cz>
26084         PR sanitize/81040
26085         * sanopt.c (rewrite_usage_of_param): New function.
26086         (sanitize_rewrite_addressable_params): Likewise.
26087         (pass_sanopt::execute): Call rewrite_usage_of_param.
26089 2017-07-03  Richard Biener  <rguenther@suse.de>
26091         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26092         back to using VIEW_CONVERT_EXPR.
26094 2017-07-03  Martin Liska  <mliska@suse.cz>
26096         PR other/78366
26097         * doc/extend.texi: Document when a resolver function is
26098         generated for target_clones.
26100 2017-07-03  Martin Liska  <mliska@suse.cz>
26102         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26103         and shadow memory of auto variables which are subject of
26104         use-after-scope sanitization.
26105         (asan_expand_mark_ifn): Add do set only when is_poison.
26107 2016-07-03  Richard Biener  <rguenther@suse.de>
26109         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26110         reduction PHIs.
26111         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26112         (vectorizable_reduction): Perform reduction PHI creation when
26113         visiting a reduction PHI and adjust and simplify code generation
26114         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26115         (vect_transform_loop): Visit reduction PHIs.
26116         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26117         defs into the SLP tree.
26118         (vect_build_slp_tree): Reduction defs terminate the recursion.
26119         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26120         of reduction defs.
26121         (vect_get_vec_defs_for_stmt_copy): Export.
26122         (vect_get_vec_defs): Likewise.
26123         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26124         purpose.
26125         (vect_get_vec_defs_for_stmt_copy): Declare.
26126         (vect_get_vec_defs): Likewise.
26128 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26130         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26131         parameter with a "loop" parameter and use it instead of the
26132         loop containing DR_STMT.  Don't check simple_iv when doing
26133         BB analysis.  Describe the two analysis modes in the comment.
26135 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26137         PR tree-optimization/69468
26138         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26139         (find_same_succ_bb): Handle ignore_edge_flags.
26141 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26143         PR tree-optimization/81192
26144         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26145         hash.
26146         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26147         differs.
26148         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26150 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26152         PR tree-optimization/81192
26153         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26154         BB_SAME_SUCC (bb) == NULL.
26156 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26158         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26159         consistency.
26161 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26163         * dumpfile.c: Include profile-count.h
26164         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26165         update profile.
26166         (insert_cond_bb): Update profile.
26167         * tree-cfg.h (insert_cond_bb): Update prototype.
26168         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26169         * tree-dump.c: Do not include tree-cfg.
26171 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26173         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26175 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26177         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26178         bb.
26180 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26182         * tree-complex.c (expand_complex_div_wide): update profile.
26184 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26185             Alan Hayward  <alan.hayward@arm.com>
26186             David Sherwood  <david.sherwood@arm.com>
26188         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26189         (CORETYPES_H): New define.
26190         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26191         (insn-modes-inline.h, s-modes-inline-h): New rules.
26192         (generated_files): Add insn-modes-inline.h.
26193         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26194         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26195         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26196         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26197         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26198         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26199         (build/gencodes.o, build/genconditions.o): Likewise.
26200         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26201         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26202         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26203         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26204         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26205         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26206         * coretypes.h: Include everything up to real.h for generators.
26207         Include insn-modes.h first.  Include wide-int-print.h after
26208         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26209         * machmode.h: Don't include insn-modes.h here.
26210         * function-tests.c: Remove includes of signop.h, machmode.h,
26211         double-int.h and wide-int.h.
26212         * rtl.h: Likewise.
26213         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26214         and wide-int.h.
26215         * optc-save-gen.awk: Likewise.
26216         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26217         * godump.c: Remove include of wide-int-print.h.
26218         * pretty-print.h: Likewise.
26219         * wide-int-print.cc: Likewise.
26220         * wide-int.cc: Likewise.
26221         * hash-map-tests.c: Remove include of signop.h.
26222         * hash-set-tests.c: Likewise.
26223         * rtl-tests.c: Likewise.
26224         * mkconfig.sh: Remove include of machmode.h.
26225         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26226         into...
26227         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26228         into an insn-modes-inline.h header file, adding appropriate
26229         include guards and end comments.
26230         (emit_insn_modes_c_header): Remove include of machmode.h.
26231         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26232         machmode.h.
26233         (main): Handle -i flag and call emit_insn_modes_inline_h when
26234         it is passed.
26236 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26238         * tree-ssa-strlen.c (strinfo): Rename the length field to
26239         nonzero_chars.  Add a full_string_p field.
26240         (compare_nonzero_chars, zero_length_string_p): New functions.
26241         (get_addr_stridx): Add an offset_out parameter.
26242         Use compare_nonzero_chars.
26243         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26244         (new_strinfo): Update after above changes to strinfo.
26245         (set_endptr_and_length): Set full_string_p.
26246         (get_string_length): Update after above changes to strinfo.
26247         (unshare_strinfo): Update call to new_strinfo.
26248         (maybe_invalidate): Likewise.
26249         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26250         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26251         as a uhwi instead of an shwi.  Update after above changes to
26252         strinfo and new_strinfo.
26253         (zero_length_string): Assert that chainsi contains full strings.
26254         Use zero_length_string_p.  Update call to new_strinfo.
26255         (adjust_related_strinfos): Update after above changes to strinfo.
26256         Copy full_string_p from origsi.
26257         (adjust_last_stmt): Use zero_length_string_p.
26258         (handle_builtin_strlen): Update after above changes to strinfo and
26259         new_strinfo.  Install the lhs as the string length if the previous
26260         entry didn't describe a full string.
26261         (handle_builtin_strchr): Update after above changes to strinfo
26262         and new_strinfo.
26263         (handle_builtin_strcpy): Likewise.
26264         (handle_builtin_strcat): Likewise.
26265         (handle_builtin_malloc): Likewise.
26266         (handle_pointer_plus): Likewise.
26267         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26268         that aren't necessarily followed by a nul terminator.
26269         (handle_char_store): Likewise.
26271 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26273         PR tree-optimization/80769
26274         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26275         for malloc and calloc.  Document the new invariant that all related
26276         strinfos have delayed lengths or none do.
26277         (verify_related_strinfos): Move earlier in file.
26278         (set_endptr_and_length): New function, split out from...
26279         (get_string_length): ...here.  Also set the lengths of related
26280         strinfos.
26281         (zero_length_string): Assert that chainsi has known (rather than
26282         delayed) lengths.
26283         (adjust_related_strinfos): Likewise.
26285 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26287         PR tree-optimization/81136
26288         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26289         assert that two references with the same misalignment have the same
26290         compile-time misalignment if those compile-time misalignments
26291         are known.
26293 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26295         * print-tree.c (print_node): Print all attributes.
26297 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26299         * cfg.c (scale_bbs_frequencies): New function.
26300         * cfg.h (scale_bbs_frequencies): Declare it.
26301         * cfgloopanal.c (single_likely_exit): Cleanup.
26302         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26303         as parameter.
26304         (scale_loop_profile): Likewise.
26305         (loop_version): Likewise.
26306         (create_empty_loop_on_edge): Update.
26307         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26308         scale_loop_frequencies, scale_loop_profile, loopify,
26309         loop_version): Update prototypes.
26310         * modulo-sched.c (sms_schedule): Update.
26311         * predict.c (unlikely_executed_edge_p): Also check probability.
26312         (probably_never_executed_edge_p): Fix typo.
26313         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26314         * tree-parloops.c (gen_parallel_loop): Update.
26315         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26316         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26317         * tree-ssa-loop-split.c (split_loop): Update.
26318         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26319         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26320         (vect_loop_versioning): Update.
26321         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26323 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26325         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26327 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26329         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26330         to keep profile consistent.
26332 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26334         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26335         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26336         * profile-count.h (max_safe_multiplier): Make unsigned.
26337         (profile_count::guessed_zero): New.
26339 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26341         * bb-reorder.c (fix_up_crossing_landing_pad,
26342         fix_crossing_conditional_branches): Use make_single_succ_edge
26343         to keep profile consistent.
26345 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26347         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26348         to update profile.
26350 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26352         PR sanitizer/81262
26353         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26354         the right scopes, make sure cond_jump isn't preserved between multiple
26355         iterations.  Search for fallthru edge whenever there are 3+ edges and
26356         use find_fallthru_edge for it.
26358 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26360         Patch by Alexander Monakov <amonakov@ispras.ru>
26361         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26362         probabilities consistently.
26364 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26366         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26367         emit_cmp_and_jump_insns.
26369 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26371         PR ipa/81261
26372         * tree-inline.c (expand_call_inline): Combine profile statuses.
26374 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26376         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26377         fold_stmt returned true.
26379 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26381         * ggc.h (empty_string): Delete.
26382         * cfgexpand.c (expand_asm_stmt): Use plain "".
26383         * optabs.c (expand_asm_memory_barrier): Likewise.
26384         * stringpool.c (empty_string): Delete.
26385         (digit_vector, digit_string): Delete.
26386         (ggc_alloc_string): Use plain "", don't optimize single digit
26387         strings.  Use ggc_alloc_atomic.
26389 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26391         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26392         comparison set and one other set, use the cost of the non-comparison
26393         set.
26395 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26397         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26398         some formatting.
26400 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26402         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26403         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26404         just set removed labels to NULL_TREE.
26406 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26408         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26409         (set_range_info): ...here.  Only call set_range_info_raw if domain
26410         is useful.
26411         (set_nonzero_bits): Call set_range_info_raw.
26412         * tree-ssanames.h (set_range_info_raw): New.
26414 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26416         PR target/81225
26417         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26418         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26419         of nonimmediate_operand and <store_mask_constraint> instead of m
26420         for the input operand.  For V8FI iterator, always split if input
26421         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26422         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26423         <store_mask_predicate> instead of register_operand and
26424         <store_mask_constraint> instead of v for the input operand.  Make
26425         sure both operands aren't MEMs for if not <mask_applied>.
26427 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26429         * lto-wrapper.c (copy_file) Close both file descriptors before
26430         exiting normally.
26432 2017-06-30  Martin Liska  <mliska@suse.cz>
26434         PR ipa/81214
26435         * multiple_target.c (create_dispatcher_calls): Make ifunc
26436         also for function that don't have calls or are not referenced.
26438 2017-06-30  Richard Biener  <rguenther@suse.de>
26440         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26441         analyze the first scalar stmt.  Move vector type computation
26442         for the BB case here from ...
26443         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26444         live operation processing in the SLP case properly.
26446 2017-06-30  Richard Biener  <rguenther@suse.de>
26448         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26450 2017-06-30  Martin Liska  <mliska@suse.cz>
26452         PR sanitizer/81021
26453         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26454         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26456 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26458         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26459         ones.
26461 2017-06-30  Richard Biener  <rguenther@suse.de>
26463         PR tree-optimization/81249
26464         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26465         condition reduction result to original scalar type.
26467 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26469         * profile-count.h (enum profile_quality): Fix typos and whitespace
26470         issues.
26472 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26474         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26475         type for branch probabilities.
26477 2017-06-29  Julian Brown  <julian@codesourcery.com>
26478             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26480         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26481         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26482         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26483         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26485 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26487         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26488         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26489         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26490         CC usage from generic code to here.
26491         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26492         CC usage into the target macros.
26494 2017-06-29  Maya Rashish  <coypu@sdf.org>
26496         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26497         objects.
26499 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26501         * arm/arm-builtins.c: Include profile-count.h
26502         * except.c (sjlj_emit_function_enter): Use
26503         profile_probability::unlikely.
26505 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26507         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26508         and tocrel_offset be pointer args rather than implicitly using
26509         static versions.
26510         (legitimate_constant_pool_address_p, rs6000_emit_move,
26511         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26512         tocrel_offset and use in toc_relative_expr_p call.
26513         (print_operand, print_operand_address): Use static tocrel_base_oac
26514         and tocrel_offset_oac.
26515         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26516         tocrel_offset_oac.
26518 2017-06-29  Maya Rashish  <coypu@sdf.org>
26520         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26522 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26524         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26525         objects, take into account only the alignment of 'op0' and 'mode1' if
26526         'op0' is a MEM.
26528 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26530         * ccmp.c (ccmp_tree_comparison_p): New function.
26531         (ccmp_candidate_p): Update to use above function.
26532         (get_compare_parts): New function.
26533         (expand_ccmp_next): Update to use new functions.
26534         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26535         new functions.
26536         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26537         take mode as argument.
26538         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26539         * expr.c (expand_expr_real_1): Pass mode as argument.
26541 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26543         * combine.c (combine_instructions): Print insns to dump_file, together
26544         with their costs.
26546 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26548         * asan.c (asan_emit_stack_protection): Update.
26549         (create_cond_insert_point): Update.
26550         * auto-profile.c (afdo_propagate_circuit): Update.
26551         * basic-block.h (struct edge_def): Turn probability to
26552         profile_probability.
26553         (EDGE_FREQUENCY): Update.
26554         * bb-reorder.c (find_traces_1_round): Update.
26555         (better_edge_p): Update.
26556         (sanitize_hot_paths): Update.
26557         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26558         (make_single_succ_edge): Update.
26559         (check_bb_profile): Update.
26560         (dump_edge_info): Update.
26561         (update_bb_profile_for_threading): Update.
26562         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26563         probabilitycount to 0.
26564         * cfgbuild.c (compute_outgoing_frequencies): Update.
26565         * cfgcleanup.c (try_forward_edges): Update.
26566         (outgoing_edges_match): Update.
26567         (try_crossjump_to_edge): Update.
26568         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26569         (expand_gimple_tailcall): Update.
26570         (construct_init_block): Use make_single_succ_edge.
26571         (construct_exit_block): Use make_single_succ_edge.
26572         * cfghooks.c (verify_flow_info): Update.
26573         (redirect_edge_succ_nodup): Update.
26574         (split_edge): Update.
26575         (account_profile_record): Update.
26576         * cfgloopanal.c (single_likely_exit): Update.
26577         * cfgloopmanip.c (scale_loop_profile): Update.
26578         (set_zero_probability): Remove.
26579         (duplicate_loop_to_header_edge): Update.
26580         * cfgloopmanip.h (loop_version): Update prototype.
26581         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26582         (force_nonfallthru_and_redirect): Update.
26583         (update_br_prob_note): Update.
26584         (rtl_verify_edges): Update.
26585         (purge_dead_edges): Update.
26586         (rtl_lv_add_condition_to_bb): Update.
26587         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26588         * cgraphunit.c (init_lowered_empty_function): Update.
26589         (cgraph_node::expand_thunk): Update.
26590         * cilk-common.c: Include profile-count.h
26591         * dojump.c (inv): Remove.
26592         (jumpifnot): Update.
26593         (jumpifnot_1): Update.
26594         (do_jump_1): Update.
26595         (do_jump): Update.
26596         (do_jump_by_parts_greater_rtx): Update.
26597         (do_compare_rtx_and_jump): Update.
26598         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26599         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26600         * dwarf2cfi.c: Include profile-count.h
26601         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26602         (sjlj_emit_dispatch_table): Likewise.
26603         * explow.c: Include profile-count.h
26604         * expmed.c (emit_store_flag_force): Update.
26605         (do_cmp_and_jump): Update.
26606         * expr.c (compare_by_pieces_d::generate): Update.
26607         (compare_by_pieces_d::finish_mode): Update.
26608         (emit_block_move_via_loop): Update.
26609         (store_expr_with_bounds): Update.
26610         (store_constructor): Update.
26611         (expand_expr_real_2): Update.
26612         (expand_expr_real_1): Update.
26613         * expr.h (try_casesi, try_tablejump): Update prototypes.
26614         * gimple-pretty-print.c (dump_probability): Update.
26615         (dump_profile): New.
26616         (dump_gimple_label): Update.
26617         (dump_gimple_bb_header): Update.
26618         * graph.c (draw_cfg_node_succ_edges): Update.
26619         * hsa-gen.c (convert_switch_statements): Update.
26620         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26621         (find_if_case_1): Update.
26622         (find_if_case_2): Update.
26623         * internal-fn.c (expand_arith_overflow_result_store): Update.
26624         (expand_addsub_overflow): Update.
26625         (expand_neg_overflow): Update.
26626         (expand_mul_overflow): Update.
26627         (expand_vector_ubsan_overflow): Update.
26628         * ipa-cp.c (good_cloning_opportunity_p): Update.
26629         * ipa-split.c (split_function): Use make_single_succ_edge.
26630         * ipa-utils.c (ipa_merge_profiles): Update.
26631         * loop-doloop.c (add_test): Update.
26632         (doloop_modify): Update.
26633         * loop-unroll.c (compare_and_jump_seq): Update.
26634         (unroll_loop_runtime_iterations): Update.
26635         * lra-constraints.c (lra_inheritance): Update.
26636         * lto-streamer-in.c (input_cfg): Update.
26637         * lto-streamer-out.c (output_cfg): Update.
26638         * mcf.c (adjust_cfg_counts): Update.
26639         * modulo-sched.c (sms_schedule): Update.
26640         * omp-expand.c (expand_omp_for_init_counts): Update.
26641         (extract_omp_for_update_vars): Update.
26642         (expand_omp_ordered_sink): Update.
26643         (expand_omp_for_ordered_loops): Update.
26644         (expand_omp_for_generic): Update.
26645         (expand_omp_for_static_nochunk): Update.
26646         (expand_omp_for_static_chunk): Update.
26647         (expand_cilk_for): Update.
26648         (expand_omp_simd): Update.
26649         (expand_omp_taskloop_for_outer): Update.
26650         (expand_omp_taskloop_for_inner): Update.
26651         * omp-simd-clone.c (simd_clone_adjust): Update.
26652         * optabs.c (expand_doubleword_shift): Update.
26653         (expand_abs): Update.
26654         (emit_cmp_and_jump_insn_1): Update.
26655         (expand_compare_and_swap_loop): Update.
26656         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26657         * predict.c (predictable_edge_p): Update.
26658         (edge_probability_reliable_p): Update.
26659         (set_even_probabilities): Update.
26660         (combine_predictions_for_insn): Update.
26661         (combine_predictions_for_bb): Update.
26662         (propagate_freq): Update.
26663         (estimate_bb_frequencies): Update.
26664         (force_edge_cold): Update.
26665         * profile-count.c (profile_count::dump): Add missing space into dump.
26666         (profile_count::debug): Add newline.
26667         (profile_count::differs_from_p): Explicitly convert to unsigned.
26668         (profile_count::stream_in): Update.
26669         (profile_probability::dump): New member function.
26670         (profile_probability::debug): New member function.
26671         (profile_probability::differs_from_p): New member function.
26672         (profile_probability::differs_lot_from_p): New member function.
26673         (profile_probability::stream_in): New member function.
26674         (profile_probability::stream_out): New member function.
26675         * profile-count.h (profile_count_quality): Rename to ...
26676         (profile_quality): ... this one.
26677         (profile_probability): New.
26678         (profile_count): Update.
26679         * profile.c (compute_branch_probabilities): Update.
26680         * recog.c (peep2_attempt): Update.
26681         * sched-ebb.c (schedule_ebbs): Update.
26682         * sched-rgn.c (find_single_block_region): Update.
26683         (compute_dom_prob_ps): Update.
26684         (schedule_region): Update.
26685         * sel-sched-ir.c (compute_succs_info): Update.
26686         * stmt.c (struct case_node): Update.
26687         (do_jump_if_equal): Update.
26688         (get_outgoing_edge_probs): Update.
26689         (conditional_probability): Update.
26690         (emit_case_dispatch_table): Update.
26691         (expand_case): Update.
26692         (expand_sjlj_dispatch_table): Update.
26693         (emit_case_nodes): Update.
26694         * targhooks.c: Update.
26695         * tracer.c (better_p): Update.
26696         (find_best_successor): Update.
26697         * trans-mem.c (expand_transaction): Update.
26698         * tree-call-cdce.c: Update.
26699         * tree-cfg.c (gimple_split_edge): Upate.
26700         (move_sese_region_to_fn): Upate.
26701         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26702         * tree-eh.c (lower_resx): Upate.
26703         (cleanup_empty_eh_move_lp): Upate.
26704         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26705         * tree-inline.c (copy_edges_for_bb): Update.
26706         (copy_cfg_body): Update.
26707         * tree-parloops.c (gen_parallel_loop): Update.
26708         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26709         (gimple_gen_time_profiler): Update.
26710         * tree-ssa-dce.c (remove_dead_stmt): Update.
26711         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26712         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26713         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26714         (unloop_loops): Update.
26715         (try_peel_loop): Update.
26716         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26717         * tree-ssa-loop-split.c (connect_loops): Update.
26718         (split_loop): Update.
26719         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26720         (hoist_guard): Update.
26721         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26722         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26723         (value_replacement): Update.
26724         * tree-ssa-reassoc.c (branch_fixup): Update.
26725         * tree-ssa-tail-merge.c (replace_block_by): Update.
26726         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26727         (create_edge_and_update_destination_phis): Update.
26728         (compute_path_counts): Update.
26729         (recompute_probabilities): Update.
26730         (update_joiner_offpath_counts): Update.
26731         (freqs_to_counts_path): Update.
26732         (duplicate_thread_path): Update.
26733         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26734         (struct switch_conv_info): Update.
26735         (gen_inbound_check): Update.
26736         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26737         (vect_do_peeling): Update.
26738         (vect_loop_versioning): Update.
26739         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26740         (optimize_mask_stores): Update.
26741         * ubsan.c (ubsan_expand_null_ifn): Update.
26742         * value-prof.c (gimple_divmod_fixed_value): Update.
26743         (gimple_divmod_fixed_value_transform): Update.
26744         (gimple_mod_pow2): Update.
26745         (gimple_mod_pow2_value_transform): Update.
26746         (gimple_mod_subtract): Update.
26747         (gimple_mod_subtract_transform): Update.
26748         (gimple_ic): Update.
26749         (gimple_stringop_fixed_value): Update.
26750         (gimple_stringops_transform): Update.
26751         * value-prof.h: Update.
26753 2017-06-29  Carl Love  <cel@us.ibm.com>
26755         * config/rs6000/rs6000-c.c: Add support for built-in functions
26756         vector signed int vec_signed (vector float);
26757         vector signed long long vec_signed (vector double);
26758         vector signed int vec_signed2 (vector double, vector double);
26759         vector signed int vec_signede (vector double);
26760         vector signed int vec_signedo (vector double);
26761         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26762         instruction generator.
26763         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26764         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26765         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26766         Add define_insn.
26767         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26768         vunsignede_v2df): Add define_expands.
26769         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26770         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26771         VEC_UNSIGNEDO): Add definitions.
26772         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26773         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26774         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26775         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26776         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26777         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26778         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26779         vec_unsignede, vec_unsignedo): Add builtin defines.
26780         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26781         declaration.
26782         * doc/extend.texi: Update the built-in documentation file for the
26783         new built-in functions.
26785 2017-06-29  Richard Biener  <rguenther@suse.de>
26787         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26788         reduction chains to LOOP_VINFO_REDUCTIONS.
26789         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26790         SLP reductions after processing reduction chains.
26792 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26794         * builtins.c (fold_builtin_FUNCTION): Use
26795         lang_hooks.decl_printable_name.
26797 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26799         PR middle-end/81194
26800         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26801         with only one label.
26802         * stmt.c (expand_case): Assert NCASES is greater than one.
26804 2017-06-29  Richard Biener  <rguenther@suse.de>
26806         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26807         anything.
26808         (group_case_labels): Likewise.
26809         (find_taken_edge): Push sanity checking on val to workers...
26810         (find_taken_edge_cond_expr): ... here
26811         (find_taken_edge_switch_expr): ... and here, handle cases
26812         with just a default label.
26813         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26814         (group_case_labels): Likewise.
26815         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26816         group_case_labels does anything cleanup the CFG again.
26818 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26820         PR tree-optimization/81196
26821         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26822         exit condition comparing two IVs.
26824 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26826         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26827         profile to the dummy entry at the end of the list of architectures.
26828         * config/arm/arm-cpu-cdata.h: Regenerated.
26830 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26831             Michael Collison <michael.collison@arm.com>
26833         PR target/70119
26834         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26835         New pattern.
26836         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26837         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26838         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26839         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26840         of shift when the shift amount is masked with constant equal to
26841         the size of the mode.
26842         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26843         predicate.
26845 2017-06-29  Martin Liska  <mliska@suse.cz>
26847         * config/i386/i386.opt: Change range from [1,5] to [0,5].
26849 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
26851         PR bootstrap/80565
26852         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26853         code.
26854         * ipa-inline.h
26855         (edge_growth_cache_entry::edge_growth_cache_entry): New
26856         function.
26857         (reset_edge_growth_cache): Update to use constructor.
26859 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26861         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26862         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26863         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26865 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
26867         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
26868         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
26870 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
26872         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
26873         (*-linux-uclibc*): Add t-uclibc tmake_file.
26874         * config/t-musl: New.
26875         * config/t-uclibc: New.
26877 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
26879         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
26880         context.
26881         (gen_comm_data): Emit architectural setting of arch_prof.
26882         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
26883         profile.
26884         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
26885         (armv8-m.base, armv8-m.main): Likewise.
26886         * arm-protos.h (arm_build_target): Add profile field.
26887         (arch_option): Likewise.
26888         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
26889         the active target.
26890         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
26891         arm_active_target.profile.
26893 2017-06-28  Richard Biener  <rguenther@suse.de>
26895         PR middle-end/81227
26896         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
26897         TYPE_OVERFLOW_WRAPS.
26898         * match.pd (negate_expr_p): Likewise.
26899         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
26900         fold_build2, not fold_binary.
26902 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26904         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
26905         Convert memory address to Pmode.
26906         (aarch64_print_operand): Assert MEM operands are always Pmode.
26908 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26910         PR target/79665
26911         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
26912         Remove redundant if.
26913         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
26914         * config/arm/aarch-common-protos.h
26915         (aarch_forward_to_shift_is_not_shifted_re): Remove.
26916         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
26918 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
26920         PR ipa/81238
26921         * multiple_target.c (create_dispatcher_calls): Set the default
26922         clone to be static, not public.
26924 2017-06-28  Richard Biener  <rguenther@suse.de>
26926         * tree-vect-loop.c (vectorizable_reduction): Move special
26927         cond reduction IV var creation ...
26928         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
26929         parameter.  Use STMT_VINFO_VECTYPE.
26930         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
26931         constant_p.
26933 2017-06-28  Martin Liska  <mliska@suse.cz>
26935         PR ipa/81128
26936         * ipa-visibility.c (non_local_p): Handle visibility.
26938 2017-06-28  Martin Liska  <mliska@suse.cz>
26940         PR driver/79659
26941         * common.opt: Add IntegerRange to various options.
26942         * opt-functions.awk (integer_range_info): New function.
26943         * optc-gen.awk: Add integer_range_info to cl_options struct.
26944         * opts-common.c (decode_cmdline_option): Handle
26945         CL_ERR_INT_RANGE_ARG.
26946         (cmdline_handle_error): Likewise.
26947         * opts.c (print_filtered_help): Show valid interval in
26948         when --help is provided.
26949         * opts.h (struct cl_option): Add range_min and range_max fields.
26950         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
26952 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
26954         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
26955         (x * C EQ/NE y * C): New transformation.
26957 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
26959         * genmultilib (combination_space): Accept '+' in option names.
26961 2017-06-28  Martin Liska  <mliska@suse.cz>
26963         PR sanitizer/81224
26964         * asan.c (instrument_derefs): Bail out inner references
26965         that are hard register variables.
26967 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
26969         PR target/81175
26970         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
26971         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
26973 2017-06-28  Richard Biener  <rguenther@suse.de>
26975         * tree-vectorizer.h (vect_get_vec_defs): Remove.
26976         (vect_get_slp_defs): Adjust.
26977         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
26978         out from ...
26979         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
26980         simplify.
26981         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
26982         get_initial_defs_for_reduction instead of vect_get_vec_defs.
26983         (vectorizable_reduction): Adjust.
26984         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
26985         handling.
26986         (vect_get_slp_defs): Likewise.
26987         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
26988         (vectorizable_bswap): Adjust.
26989         (vectorizable_call): Likewise.
26990         (vectorizable_conversion): Likewise.
26991         (vectorizable_assignment): Likewise.
26992         (vectorizable_shift): Likewise.
26993         (vectorizable_operation): Likewise.
26994         (vectorizable_store): Likewise.
26995         (vectorizable_condition): Likewise.
26996         (vectorizable_comparison): Likewise.
26998 2017-06-28  Michael Collison  <michael.collison@arm.com>
27000         PR target/68535
27001         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27002         set of base_reg
27003         (arm_gen_movmemqi): Removed unused variable 'i'.
27004         Convert 'for' loop into 'while' loop.
27005         (arm_expand_prologue): Remove last unnecessary set of insn.
27006         (thumb_pop): Remove unused variable 'pushed_words'.
27007         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27009 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27011         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27012         * config/s390/s390-protos.h: Add prototype of
27013         s390_rel_address_ok_p.
27014         * config/s390/s390.c (s390_got_symbol): New function.
27015         (s390_rel_address_ok_p): New function.
27016         (legitimize_pic_address): Use s390_rel_address_ok_p.
27017         (s390_load_got): Use s390_got_symbol.
27018         (s390_option_override): Issue error if
27019         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27020         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27021         New macro.
27022         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27024 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27026         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27027         (X * copysign (1.0, X)): New pattern.
27028         (X * copysign (1.0, -X)): New pattern.
27029         (copysign (-1.0, CST)): New pattern.
27031 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27033         * genmultilib (combination_space): Remove variable.
27034         Validate reuse rules against regular expression for any sequence
27035         of multilib options in any order.
27037 2017-06-27  Michael Collison  <michael.collison@arm.com>
27039         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27040         call aarch64_split_simd_combine.
27041         * (aarch64_combine_internal<mode>): Delete pattern.
27042         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27043         Allow register and subreg operands.
27045 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27047         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27048         specific need, just fallback on defaults.
27049         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27051 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27052             Olivier Hainque  <hainque@adacore.com>
27054         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27055         map for 64bits.
27056         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27057         targets. Pick a default if no particular attempt applied.
27058         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27059         larger contexts.
27061 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27063         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27064         (x86_64-wrs-vxworks7): Likewise.
27066 2017-06-27  Marek Polacek  <polacek@redhat.com>
27068         PR sanitizer/81223
27069         * ubsan.c (instrument_null): Check get_base_address's result for null.
27071 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27073         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27075 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27077         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27078         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27079         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27080         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27081         New function types.
27082         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27083         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27084         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27085         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27086         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27087         BUILT_IN_FEUPDATEENV): New builtins.
27088         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27089         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27090         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27091         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27092         macros.
27093         (builtin_structptr_types): Adjust size.
27094         * tree.c (builtin_structptr_types): Add four entries.
27096 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27097             Olivier Hainque  <hainque@adacore.com>
27099         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27100         (TLS_SYM): New local macro, forcing reference to __tls__ on
27101         link command lines for VxWorks 7 RTPs, triggering initialization
27102         of tlsLib.
27103         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27104         OS features TLS support, true for RTPs on VxWorks 7.
27105         * config/vxworks.c (vxworks_override_options): Setup emutls
27106         accordingly.
27108 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27110         * predict.c (test_prediction_value_range): Use -1U instead of -1
27111         to avoid narrowing conversion warning.
27112         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27113         to avoid narrowing conversion warning.
27114         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27115         -1.
27116         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27118 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27120         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27121         64bit configurations.
27122         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27123         (SIZE_TYPE): Likewise.
27124         * config/vxworks.c (vxworks_emutls_var_fields): Use
27125         long_unsigned_type_node instead of unsigned_type_node as the offset
27126         field type, which is "pointer" mode in emutls.c.
27128 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27130         PR sanitizer/81209
27131         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27133         PR middle-end/81207
27134         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27135         gimple_vuse copying separately from gimple_vdef copying.
27137 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27139         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27140         condition.
27142 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27143             Olivier Hainque  <hainque@adacore.com>
27145         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27146         for all vxworks7 targets.
27147         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27148         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27149         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27150         variations for VX6/VX7 and 32/64bits later on in ...
27151         (VXWORKS_LIB_SPEC): Leverage new macros.
27152         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27153         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27155 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27157         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27158         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27160 2017-06-26  Carl Love  <cel@us.ibm.com>
27162         * config/rs6000/rs6000-c.c: Add support for built-in functions
27163         vector bool char vec_reve (vector bool char);
27164         vector signed char vec_reve (vector signed char);
27165         vector unsigned char vec_reve (vector unsigned char);
27166         vector bool int vec_reve (vector bool int);
27167         vector signed int vec_reve (vector signed int);
27168         vector unsigned int vec_reve (vector unsigned int);
27169         vector bool long long vec_reve (vector bool long long);
27170         vector signed long long vec_reve (vector signed long long);
27171         vector unsigned long long vec_reve (vector unsigned long long);
27172         vector bool short vec_reve (vector bool short);
27173         vector signed short vec_reve (vector signed short);
27174         vector double vec_reve (vector double);
27175         vector float vec_reve (vector float);
27176         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27177         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27178         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27179         (altivec_vreve): New pattern.
27180         * config/rs6000/altivec.h (vec_reve): New define.
27181         * doc/extend.texi (vec_rev): Update the built-in documentation file
27182         for the new built-in functions.
27184 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27186         PR tree-optimization/71815
27187         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27188         function.
27189         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27190         has_single_use.
27191         (slsr_process_phi): Likewise.
27192         (replace_uncond_cands_and_profitable_phis): Don't replace a
27193         multiply candidate with a stride of 1 (copy or cast).
27194         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27195         has_single_use.
27196         (lowest_cost_path): Likewise.
27197         (total_savings): Likewise.
27199 2017-06-26  Richard Biener  <rguenther@suse.de>
27201         PR target/81175
27202         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27203         Use def_builtin_pure for all gather builtins.
27205 2017-06-26  Richard Biener  <rguenther@suse.de>
27207         PR tree-optimization/81203
27208         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27209         non-dominating BBs.
27211 2017-06-26  Marek Polacek  <polacek@redhat.com>
27213         PR c/80116
27214         * doc/invoke.texi: Document -Wmultistatement-macros.
27216 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27218         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27219         arm_neon_ok_no_float_abi effective target.
27221 2017-06-26  Richard Biener  <rguenther@suse.de>
27223         PR tree-optimization/80928
27224         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27225         (copy_bbs): Set BB_DUPLICATED flag early.
27226         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27227         marked blocks.
27228         (execute_on_shrinking_pred): Likewise.
27229         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27230         BB_DUPLICATED blocks.
27231         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27232         iterate over all PHIs considering removal of *gsi.
27234 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27236         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27237         qdf24xx.
27239 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27241         * config/rs6000/rs6000-string.c: (expand_block_clear,
27242         do_load_for_compare, select_block_compare_mode,
27243         compute_current_alignment, expand_block_compare,
27244         expand_strncmp_align_check, expand_strn_compare,
27245         expand_block_move, rs6000_output_load_multiple)
27246         Move functions related to string/block move/compare
27247         to a separate file.
27248         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27249         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27250         for this function which is now used in two files.
27251         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27252         * config.gcc: Add rs6000-string.o to extra_objs for
27253         targets powerpc*-*-* and rs6000*-*-*.
27255 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27257         PR target/80510
27258         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27259         32-bit, since indexed is not valid for DImode.
27260         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27261         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27262         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27263         (define_peephole2 for Altivec d-form store): Likewise.
27265         PR ipa/81185
27266         * multiple_target.c (create_dispatcher_calls): Only create the
27267         dispatcher call if the function is the default clone of a
27268         versioned function.
27270 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27272         PR middle-end/80902
27273         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27274         a call, force the call to not be a tail call.
27276 2017-06-23  Jeff Law  <law@redhat.com>
27278         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27279         enquire.
27281 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27283         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27284         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27285         handling for early expansion of vector shifts (sl,sr,sra,rl).
27286         (builtin_function_type): Add vector shift right instructions
27287         to the unsigned argument list.
27289 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27291         rtl-optimizatoin/79286
27292         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27293         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27294         trap.  PIC register plus a const unspec without offset can never trap.
27296 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27298         * tree.h (builtin_structptr_type): New type.
27299         (builtin_structptr_types): Declare new array.
27300         * tree.c (builtin_structptr_types): New array.
27301         (free_lang_data, build_common_tree_nodes): Use it.
27303 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27305         PR c++/81187
27306         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27307         -Wnoexcept.
27309 2017-06-22  Matt Turner  <mattst88@gmail.com>
27311         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27312         Lake models to skylake case.  Assume skylake for unknown
27313         models with clflushopt.
27315 2017-06-22  Jeff Law  <law@redhat.com>
27317         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27318         frame sizes that do not satisfy aarch64_uimm12_shift.
27320 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27322         * profile-count.h (apply_probability,
27323         apply_scale, probability_in): Fix checks for zero.
27325 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27327         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27328         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27330 2016-06-22  Richard Biener  <rguenther@suse.de>
27332         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27333         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27334         REDUC_MAX_EXPR support.
27335         (vectorizable_reduction): Likewise.
27336         (vect_create_epilog_for_reduction): Likewise.
27338 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27340         * match.pd (A / (1 << B) -> A >> B): New.
27341         * generic-match-head.c: Include optabs-tree.h.
27342         * gimple-match-head.c: Likewise.
27343         * optabs-tree.h (target_supports_op_p): New.
27344         * optabs-tree.c (target_supports_op_p): New.
27346 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27348         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27349         $gcc_cv_ld --help output.
27350         (gcc_cv_ld_demangle): Likewise.
27351         (gcc_cv_ld_eh_frame_hdr): Likewise.
27352         (gcc_cv_ld_pie): Likewise.
27353         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27354         (gcc_cv_ld_buildid): Likewise.
27355         (gcc_cv_ld_sysroot): Likewise.
27356         (ld_bndplt_support): Likewise.
27357         (ld_pushpopstate_support): Likewise.
27358         * configure: Regenerate.
27359         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27361 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27363         PR target/81151
27364         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27365         operands indexes to avoid gap between operands and match_dups.
27367 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27369         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27370         Increment Arith_shift and Arith_shift_reg by 1.
27371         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27372         New tuning flag.
27373         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27374         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27375         (aarch64_strip_extend): Add new argument and test for it.
27376         (aarch64_cheap_mult_shift_p): New function.
27377         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27378         add a cost if it is true.
27379         Update calls to aarch64_strip_extend.
27380         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27382 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27384         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27385         tunings.
27386         (thunderxt88): Likewise.
27387         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27388         (thunderx_prefetch_tune): New variable.
27389         (thunderx2t99_prefetch_tune): Update for the correct values.
27390         (thunderxt88_tunings): New variable.
27391         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27392         generic_prefetch_tune.
27393         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27395 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27397         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27398         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27399         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27400         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27401         (aarch64_atomic_cas<mode>, GPI): Likewise.
27403 2017-06-21  Martin Liska  <mliska@suse.cz>
27405         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27406         statements on cold and hot labels.
27407         * predict.c (tree_estimate_probability_bb): Remove the
27408         prediction from this place.
27410 2017-06-21  Martin Liska  <mliska@suse.cz>
27412         PR tree-optimization/79489
27413         * gimplify.c (maybe_add_early_return_predict_stmt): New
27414         function.
27415         (gimplify_return_expr): Call the function.
27416         * predict.c (tree_estimate_probability_bb): Remove handling
27417         of early return.
27418         * predict.def: Update comment about early return predictor.
27419         * gimple-predict.h (is_gimple_predict): New function.
27420         * predict.def: Change default value of early return to 66.
27421         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27422         statements.
27423         * passes.def: Put pass_strip_predict_hints to the beginning of
27424         IPA passes.
27426 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27428         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27429         FUNCTION_DECL declarations.
27430         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27431         declarations.
27432         (dwaf2out_decl): Likewise.
27433         * godump.c (go_early_global_decl): Skip call to the real debug hook
27434         for FUNCTION_DECL declarations.
27435         * passes.c (rest_of_decl_compilation): Skip call to the
27436         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27437         -fdump-go-spec is passed.
27439 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27441         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27442         Reorder for compactness.
27443         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27444         (def_builtin_pure, def_builtin_pure2): New functions.
27445         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27447 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27449         * match.pd (nop_convert): New predicate.
27450         ((A +- CST1) +- CST2): Allow some NOP conversions.
27452 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27454         PR c++/81130
27455         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27456         with ctors/dtors if GOVD_SHARED is set.
27458 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27460         * config/aarch64/aarch64.md (movti_aarch64):
27461         Emit mov rather than orr.
27462         (movtf_aarch64): Likewise.
27463         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27464         Emit mov rather than orr.
27466 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27468         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27469         Swap alternatives, make integer dup more expensive.
27471 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27473         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27474         Return true for non-tls symbols.
27476 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27478         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27479         (cortex-a75): Likewise.
27480         (cortex-a75.cortex-a55): Likewise.
27481         * config/aarch64/aarch64-tune.md: Regenerate.
27482         * doc/invoke.texi (-mtune): Document new values for -mtune.
27484 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27486         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27487         stack_size feature.
27488         (Effective-Target Keywords, Other attributes): Suggest using
27489         dg-add-options stack_size feature to get stack limit in stack_size
27490         effective target documentation.
27492 2017-06-21  Julian Brown  <julian@codesourcery.com>
27493             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27495         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27496         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27497         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27498         reservation.
27499         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27500         attribute type list for neon_multiply.
27501         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27502         attribute type list for neon_multiply.
27503         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27504         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27505         attribute type list for neon_multiply.
27506         * config/arm/types.md (crypto_pmull): Add.
27507         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27508         attribute type list.
27510 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27512         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27513         arm1176jzf-s.
27515 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27517         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27518         to make sure not to dereference a NULL cost_classes_ptr pointer.
27520 2017-06-20  Carl Love  <cel@us.ibm.com>
27522         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27523         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27524         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27525         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27526         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27527         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27528         VMULOSW): New enum "unspec" values.
27529         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27530         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27531         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27532         altivec_vmulosw): New patterns.
27533         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27534         VMULOSW): Add definitions.
27536 2017-06-20  Julia Koval  <julia.koval@intel.com>
27538         * config/i386/i386.c: Fix rounding expand for new pattern.
27539         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27541 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27543         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27544         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27546 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27548         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27549         feature string.
27551 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27553         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27554         architecture, then by implementer ID.
27555         * config/aarch64/aarch64-tune.md: Regenerate.
27557 2017-06-20  Richard Biener  <rguenther@suse.de>
27559         PR middle-end/81097
27560         * fold-const.c (split_tree): Fold to type before negating.
27562 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27564         * diagnostic-show-locus.c
27565         (selftest::test_fixit_deletion_affecting_newline): New function.
27566         (selftest::diagnostic_show_locus_c_tests): Call it.
27568 2017-06-20  Andreas Schwab  <schwab@suse.de>
27570         PR target/80970
27571         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27572         instead of "+d".
27574 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27576         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27577         __ARM_FEATURE_COPROC according to support.
27579 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27581         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27582         Rewritten to avoid overflow for > 32-bit pointers.
27584         PR sanitizer/81125
27585         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27586         by removing enum keyword.
27587         (ubsan_type_descriptor): Likewise.  Formatting fix.
27589         PR target/81121
27590         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27591         splitter): Require TARGET_SSE2 in the condition.
27593 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27595         PR target/79799
27596         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27597         for doing vector set of SFmode on ISA 3.0.
27598         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27599         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27600         element.
27601         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27602         SFmode value into a V4SF variable that was extracted from another
27603         V4SF variable without converting the element to double precision
27604         and back to single precision vector format.
27605         (vsx_insert_extract_v4sf_p9_2): Likewise.
27607 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27609         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27610         in UWHI to avoid undefined overflow.
27612         PR sanitizer/81125
27613         * ubsan.h (enum ubsan_encode_value_phase): New.
27614         (ubsan_encode_value): Change second argument to
27615         enum ubsan_encode_value_phase with default value of
27616         UBSAN_ENCODE_VALUE_GENERIC.
27617         * ubsan.c (ubsan_encode_value): Change second argument to
27618         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27619         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27620         create_tmp_var_raw instead of create_tmp_var and use a
27621         TARGET_EXPR.
27622         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27623         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27624         ubsan_encode_value callers.
27626         PR sanitizer/81111
27627         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27628         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27629         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27631 2017-06-19  Richard Biener  <rguenther@suse.de>
27633         PR middle-end/81118
27634         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27635         estimates if we changed anything.
27637 2017-06-19  Richard Biener  <rguenther@suse.de>
27639         PR tree-optimization/80887
27640         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27641         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27642         simplified lookups, then reset mprts_hook.
27643         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27644         simplifying.
27645         (try_to_simplify): Likewise.
27647 2017-06-19  Martin Liska  <mliska@suse.cz>
27649         PR sanitizer/80879
27650         * gimplify.c (gimplify_switch_expr):
27651         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27653 2017-06-19  Martin Liska  <mliska@suse.cz>
27655         * doc/install.texi: Document that PGO runs in 4 stages.
27657 2017-06-19  Martin Liska  <mliska@suse.cz>
27659         PR ipa/80732
27660         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27661         to dispatcher function name.
27662         * multiple_target.c (replace_function_decl): New function.
27663         (create_dispatcher_calls): Redirect both edges and references.
27665 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27667         * profile-count.c (profile_count::dump): Dump quality.
27668         (profile_count::differs_from_p): Update for unsigned val.
27669         * profile-count.h (profile_count_quality): New enum.
27670         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27672 2017-06-19  Richard Biener  <rguenther@suse.de>
27674         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27675         struct function as arg.
27676         (estimate_numbers_of_iterations): Export overload with loop arg.
27677         (free_numbers_of_iterations_estimates_loop): Use an overload of
27678         free_numbers_of_iterations_estimates instead.
27679         * tree-cfg.c (remove_bb): Adjust.
27680         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27681         * tree-parloops.c (gen_parallel_loop): Likewise.
27682         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27683         Likewise.
27684         (tree_unroll_loops_completely): Likewise.
27685         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27686         Use an overload instead and export.
27687         (estimated_loop_iterations): Adjust.
27688         (max_loop_iterations): Likewise.
27689         (likely_max_loop_iterations): Likewise.
27690         (estimate_numbers_of_iterations): Take struct function as arg
27691         and adjust.
27692         (loop_exits_before_overflow): Adjust.
27693         (free_numbers_of_iterations_estimates_loop): Use an overload.
27694         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27695         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27697 2017-06-19  Richard Biener  <rguenther@suse.de>
27699         PR ipa/81112
27700         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27701         RANGE_EXPR conservatively.
27703 2017-06-16  Carl Love  <cel@us.ibm.com>
27705         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27706         definitions for vec_float, vec_float2, vec_floato,
27707         vec_floate built-ins.
27708         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27709         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27710         floate.
27711         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27712         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27713         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27714         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27715         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27716         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27717         vec_floato): Add builtin defines.
27718         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27719         Update the built-in documentation file for the new built-in
27720         functions.
27722 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27724         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27725         (mthumb): Mark as the negative of -marm.
27727 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27729         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27730         extension options.
27731         (ARM Options, -mtune): Document that this accepts the same
27732         extension options as -mcpu.
27733         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27735 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27737         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27738         permitted extensions.
27740 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27742         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27743         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27744         (armv8-m.main): Add option +nodsp.
27745         * config/arm/arm-cpu-cdata.h: Regenerated.
27747 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27749         * config/arm/t-fuchsia: New file.
27750         * config.gcc (arm*-*-fuchsia*): Use it.
27752 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27754         * config/arm/t-symbian: Rewrite for new option infrastructure.
27756 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27758         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27759         (MULTILIB_REQUIRED): Likewise.
27761 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27763         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27764         (MULTILIB_RESUE): Likewise.
27765         (MULTILIB_MATCHES): Likewise.
27766         (MULTLIB_REQUIRED): Likewise.
27768 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27770         * config/arm/t-rtems: Rewrite for new option framework.
27772 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27774         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27775         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27776         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27777         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27778         * config/arm/t-multilib: ... here.
27779         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27780         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27781         armv7-a and armv8*-a when A-profile libraries have not been built.
27782         * config/arm/t-rmprofile: Rewrite.
27784 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27786         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27787         with a backslash.  Remove the backslash after substituting unescaped
27788         periods.
27789         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27791 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27793         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27794         the driver to pass through the default setting of -mfloat-abi.
27795         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27796         rather than NULL.
27797         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27798         (all_feat_combs): New rule.
27799         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27800         default libraries.
27801         * config/arm/t-aprofile: Rewrite.
27803 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27805         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27806         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27807         fpu is not specified by the user/command-line.
27808         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27809         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27810         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27811         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27812         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27813         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27815 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27817         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27818         * config/arm/t-arm-elf: Rewritten.
27820 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27822         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27823         have some floating-point instructions.
27824         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27825         (TARGET_MAYBE_HARD_FLOAT): New macro.
27826         * config/arm/arm-builtins.c (arm_init_builtins): Use
27827         TARGET_MAYBE_HARD_FLOAT.
27828         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27830 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27832         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27833         (configargs.h): Include it.
27834         (arm_print_hint_for_fpu_option): New function.
27835         (arm_parse_fpu_option): New function.
27836         (candidate_extension): New class.
27837         (arm_canon_for_multilib): New function.
27838         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27839         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27840         (ARCH_CANONICAL_SPECS): New macro.
27841         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27843 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27845         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27846         are set after handling multilib fragments.  Set target_cpu_default2
27847         from with_cpu.
27849 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27851         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27852         cpu name.
27853         (arm*-*-*): Set target_cpu_default2 to a quoted string.
27854         * config/arm/parsecpu.awk (check_cpu): Validate any extension
27855         options.
27856         (check_arch): Likewise.
27857         * config/arm/arm.c (arm_configure_build_target): Handle
27858         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
27859         options in the default.
27861 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27863         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
27864         when an option is an alias of another.
27865         * config/arm/parsecpu.awk (optalias): New parser token.
27866         (gen_comm_data): Mark non-alias options as such.  Emit entries
27867         for extension aliases.
27868         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
27869         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
27870         (armv6kz, armv6zk, armv6t2): Likewise.
27871         (armv7): Make vfpv3-d16 an alias.
27872         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
27873         canonical order.
27874         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
27875         Sort in canonical order.
27876         (armv8-a): Sort in canonical order.
27877         (armv8.1-a, armv8.2-a):  Likewise.
27878         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
27879         canonical order.
27880         (cortex-a9): Sort in canonical order.
27881         * config/arm/arm.c (selftests.h): Include it.
27882         (arm_test_cpu_arch_data): New function.
27883         (arm_run_self_tests): New function.
27884         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
27885         (targetm): Move declaration to the end of the file.
27886         * arm-cpu-cdata.h: Regenerated.
27888 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27890         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
27891         call to target_mode_check describing the type of option passed.
27892         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
27893         (arm_target_thumb_only): Use arm_parse_arch_option_name or
27894         arm_parse_cpu_option_name to match parameters against list of
27895         available targets.
27896         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
27897         arm_arch_core_flags data structure.
27898         * config/arm/arm-cpu_cdata.h: Regenerated.
27900 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27902         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
27903         config/arm/arm.c.
27904         (arm_print_hint_for_cpu_option): Likewise.
27905         (arm_print_hint_for_arch_option): Likewise.
27906         (arm_parse_cpu_option_name): Likewise.
27907         (arm_parse_arch_option_name): Likewise.
27908         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
27909         of entries in the all_fpus list.
27910         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
27911         (arm_parse_cpu_option_name): Declare.
27912         (arm_parse_arch_option_name): Declare.
27913         (arm_parse_option_features): Declare.
27914         (arm_intialize_isa): Declare.
27915         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
27916         data tables to ...
27917         (gen_comm_data): ... here.  Make definitions non-static.
27918         * config/arm/arm-cpu-data.h: Regenerated.
27919         * config/arm/arm-cpu-cdata.h: Regenerated.
27921 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27923         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
27924         (cpu_arch_extension): New structure.
27925         (cpu_arch_option, arch_option, cpu_option): New structures.
27926         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
27927         architecture types.
27928         (gen_data): Generate new format data tables.
27929         * config/arm/arm.c (cpu_tune): New structure.
27930         (cpu_option, processors): Delete.
27931         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
27932         (arm_print_hint_for_cpu_option): ... this and ...
27933         (arm_print_hint_for_arch_option): ... this.
27934         (arm_parse_arch_cpu_name): Delete.  Replace with ...
27935         (arm_parse_cpu_option_name): ... this and ...
27936         (arm_parse_arch_option_name): ... this.
27937         (arm_unrecognized_feature): Change type of target parameter to
27938         cpu_arch_option.
27939         (arm_parse_arch_cpu_features): Delete.  Replace with ...
27940         (arm_parse_option_features): ... this.
27941         (arm_configure_build_target): Rework to use new configuration data
27942         tables.
27943         (arm_print_tune_info): Rework for new configuration data tables.
27944         * config/arm/arm-cpu-data.h: Regenerated.
27945         * config/arm/arm-cpu.h: Regenerated.
27947 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27949         * Makefile.in (OBJS): Move sbitmap.o from here ...
27950         (OBJS-libcommon): ... to here.
27952 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27954         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
27955         (ISA_ALL_CRYPTO): New macro.
27956         (ISA_ALL_SIMD): New macro
27957         (ISA_ALL_FP): New macro.
27958         * config/arm/arm.c (fpu_bitlist): Update initializer.
27959         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
27960         simd or fp.
27961         (arm9e): Add fpu.  Add option for nofp
27962         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
27963         (arm926ej-s, arm1026ej-s): Likewise.
27964         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
27965         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
27966         neon-fp16, neon-vfpv4, nofp and nosimd.
27967         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
27968         (cortex-a8): Add fpu.  Add option for nofp.
27969         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
27970         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
27971         (cortex-r4f): Add fpu.
27972         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
27973         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
27974         for nofp.
27975         (cortex-r8): Likewise.
27976         (cortex-m4): Add fpu.  Add option for nofp.
27977         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
27978         (cortex-a17.cortex-a7): Likewise.
27979         (cortex-a32): Add fpu.  Add options for crypto and nofp.
27980         (cortex-a35, cortex-a53): Likewise.
27981         (cortex-a57): Add fpu.  Add option for crypto.
27982         (cortex-a72, cortex-a73): Likewise.
27983         (exynos-m1): Likewise.
27984         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
27985         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
27986         (cortex-m33): Add fpu.  Add option for nofp.
27987         * config/arm/arm-cpu-cdata.h: Regenerated
27988         * config/arm/arm-cpu-data.h: Regenerated.
27990 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27992         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
27993         (armv5te, armv5tej): Likewise.
27994         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
27995         (armv7): Add options fp and vfpv3-d16.
27996         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
27997         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
27998         nofp and nosimd.
27999         (armv7ve): Likewise.
28000         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28001         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28002         (armv8-a): Add nocrypto option.
28003         (armv8.1-a, armv8.2-a): Likewise.
28004         (armv8-m.main): add options fp, fp.dp and nofp.
28006 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28008         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28009         nofp.
28010         (armv8-a+crc): Delete.
28011         (armv8.1-a): Add options simd, crypto and nofp.
28012         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28013         (armv8.2-a+fp16): Delete.
28014         (armv8-m.main): Add option dsp.
28015         (armv8-m.main+dsp): Delete.
28016         (cortex-a8): Add fpu.  Add nofp option.
28017         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28018         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28019         link to main cpu and architecture data structures.
28020         (gen_comm_data): Only put isa attributes from the main architecture
28021         in common tables.
28022         (option): New statement for architecture and CPU entries.
28023         * arm.c (struct cpu_option): New structure.
28024         (struct processors): Add entry for options.
28025         (arm_unrecognized_feature): New function.
28026         (arm_parse_arch_cpu_name): Ignore any characters after the first
28027         '+' character.
28028         (arm_parse_arch_cpu_feature): New function.
28029         (arm_configure_build_target): Separate out any CPU and architecture
28030         features and parse separately.  Don't error out if -mfpu=auto is
28031         used with only an architecture string.
28032         (arm_print_asm_arch_directives): New function.
28033         (arm_file_start): Call it.
28034         * config/arm/arm-cpu-cdata.h: Regenerated.
28035         * config/arm/arm-cpu-data.h: Likewise.
28036         * config/arm/arm-tables.opt: Likewise.
28038 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28040         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28041         assembler when it is not -mfpu=auto.
28043 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28045         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28046         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28047         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28048         (ASM_CPU_SPEC): Rewrite.
28049         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28050         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28051         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28052         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28053         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28054         copied string is NUL-terminated.  Also strip any characters prefixed
28055         by '+'.
28056         (arm_rewrite_selected_arch): New function.
28057         (arm_rewrite_march): New function.
28059 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28061         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28062         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28063         (march, mcpu, mtune): Convert to string-based options.
28064         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28065         (arm_parse_arch_cpu_name): New function.
28066         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28067         identify selected architecture or CPU.
28068         (arm_option_save): New function.
28069         (TARGET_OPTION_SAVE): Redefine.
28070         (arm_option_restore): Restore string options.
28071         (arm_option_print): Print string options.
28073 2017-06-16  Martin Sebor  <msebor@redhat.com>
28075         PR tree-optimization/80933
28076         PR tree-optimization/80934
28077         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28078         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28079         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28080         (gimple_fold_builtin): Call them.
28082 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28084         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28085         as unlikely; update profile.
28087 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28089         * predict.c (force_edge_cold): Handle declaring edges impossible
28090         more aggresively.
28092 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28094         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28095         profile.
28096         (try_unroll_loop_completely): Fix reporting.
28098 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28100         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28102 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28104         PR target/71778
28105         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28106         if given a non-constant argument for an intrinsic which requires a
28107         constant.
28109 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28111         * profile.c (compare_freqs): New function.
28112         (branch_prob): Sort edge list.
28113         (find_spanning_tree): Assume that the list is priority sorted.
28115 2017-06-16  Richard Biener  <rguenther@suse.de>
28117         PR tree-optimization/81090
28118         * passes.def (pass_record_bounds): Remove.
28119         * tree-pass.h (make_pass_record_bounds): Likewise.
28120         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28121         make_pass_record_bounds): Likewise.
28122         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28123         not free niter estimates at the beginning but at the end.
28124         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28126 2017-06-16  Richard Biener  <rguenther@suse.de>
28128         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28129         initializer to workaround ICE in host GCC 4.8.
28131 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28133         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28134         counts.
28135         (clone_inlined_nodes): Update.
28137 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28139         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28140         prefetch settings, and enable prefetching by default at -O3.
28142 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28144         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28145         Set flag_prefetch_loop_arrays according to tuning data.
28147 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28149         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28150         New tune structure.
28151         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28152         [Unrelated to main purpose of the patch] Place the pointer field last
28153         to enable type checking errors when tune structure are wrongly merged.
28154         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28155         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28156         (thunderx2t99_prefetch_tune): New tune constants.
28157         (tune_params *_tunings): Update all tunings (no functional change).
28158         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28159         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28160         from tunings structures.
28162 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28164         PR sanitizer/81094
28165         * ubsan.c (instrument_null): Add T argument, use it instead
28166         of computing it based on IS_LHS.
28167         (instrument_object_size): Likewise.
28168         (pass_ubsan::execute): Adjust instrument_null and
28169         instrument_object_size callers to pass gimple_get_lhs or
28170         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28171         calling get_base_address and instrument_mem_ref.  Handle
28172         aggregate call arguments for object-size sanitization.
28174 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28176         PR tree-optimization/81089
28177         * tree-vrp.c (is_masked_range_test): Validate operands of
28178         subexpression.
28180 2017-06-15  Martin Sebor  <msebor@redhat.com>
28182         PR c++/80560
28183         * dumpfile.c (dump_register): Avoid calling memset to initialize
28184         a class with a default ctor.
28185         * gcc.c (struct compiler): Remove const qualification.
28186         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28187         * hash-table.h: Ditto.
28188         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28189           assignment.
28190         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28191         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28192         default ctor.
28193         * params.h (struct param_info): Make struct members non-const.
28194         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28195         with default initialization.
28196         * vec.h (vec_copy_construct, vec_default_construct): New helper
28197         functions.
28198         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28199         with vec_copy_construct.
28200         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28201         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28202         * doc/invoke.texi (-Wclass-memaccess): Document.
28204 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28206         * emit-rtl.h (is_leaf): Update comment about local
28207         register allocator.
28209 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28211         PR target/78818
28212         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28213         for a variable to have a section before checking if the section has a
28214         name.
28215         Set section to.persistent if persistent attribute is set.
28216         Warn if .persistent attribute is used on an automatic variable.
28218 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28220         PR rtl-optimization/80474
28221         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28223 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28225         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28226         of REGNO.
28228 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28230         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28231         (casesi): Emit bounds checking as RTL.
28232         (casesi_internal_mips16_<mode>): Remove bounds checking.
28234 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28236         * config/xtensa/xtensa.c (xtensa_option_override): Append
28237         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28238         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28239          xtensa_doloop_hooks): Define unconditionally.
28240         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28241         TARGET_LOOPS.
28242         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28243         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28244         for it in xtensa_option_override.
28245         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28246          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28248 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28250         * doc/cppopts.texi: Document '-' special value to -MF.
28252 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28254         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28255         (cortex_a53_fconst): Likewise.
28256         (cortex_a53_fpmul): Likewise.
28257         (cortex_a53_f_load_64): Likewise.
28258         (cortex_a53_f_load_many): Likewise.
28259         (cortex_a53_advsimd_alu): Likewise.
28260         (cortex_a53_advsimd_alu_q): Likewise.
28261         (cortex_a53_advsimd_mul): Likewise.
28262         (cortex_a53_advsimd_mul_q): Likewise.
28263         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28264         Add missing fmul, r2f_cvt and fconst cases.
28266 2017-06-14  Richard Biener  <rguenther@suse.de>
28268         PR middle-end/81088
28269         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28270         literal constants.
28271         (fold_binary_loc): When associating do not treat pre-existing
28272         TREE_OVERFLOW on literal constants as a reason to allow
28273         TREE_OVERFLOW on associated literal constants.
28275 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28277         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28278         (MASK_FEATURES): New macro.
28279         * config/sparc/sparc.c (sparc_option_override): Remove the special
28280         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28282 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28284         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28285         a division of 0 if non-call exceptions are enabled.
28287 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28288             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28290         PR target/71663
28291         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28292         Improve vector initialization code gen for only variable case.
28294 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28296         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28298 2017-06-14  Richard Biener  <rguenther@suse.de>
28300         PR tree-optimization/81083
28301         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28302         as values.
28304 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28306         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28307         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28308         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28309         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28310         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28311         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28313 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28315         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28316         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28318 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28320         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28322 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28324         * config/rs6000/t-rtems: Don't handle SPE.
28326 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28328         * config/rs6000/t-linux: Don't handle SPE.
28330 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28332         * config/rs6000/eabispe.h: Delete file.
28334 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28336         * config/rs6000/t-spe: Delete file.
28338 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28340         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28341         (rs6000_legitimate_offset_address_p): Return false for anything in
28342         V2SImode or V2SFmode.
28344 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28346         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28347         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28348         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28349         and V4HImode.
28350         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28351         (rs6000_legitimate_offset_address_p): Ditto.
28352         (rs6000_emit_move): Ditto.
28353         (rs6000_init_builtins): Remove V4HI_type_node.
28355 2017-06-13  Martin Liska  <mliska@suse.cz>
28357         PR sanitize/78204
28358         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28359         (gate_asan): Likewise.
28360         * asan.h (asan_no_sanitize_address_p): Remove the function.
28361         (sanitize_flags_p): New function.
28362         * builtins.def: Fix coding style.
28363         * common.opt: Use renamed enum value.
28364         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28365         * doc/extend.texi: Document no_sanitize attribute.
28366         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28367         to SANITIZE_UNDEFINED_NONDEFAULT.
28368         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28369         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28370         Use sanitize_flags_p.
28371         * gimplify.c (gimplify_function_tree): Likewise.
28372         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28373         * opts.c (parse_no_sanitize_attribute): New function.
28374         (common_handle_option): Use renamed enum value.
28375         * opts.h (parse_no_sanitize_attribute): Declare.
28376         * tree.c (sanitize_flags_p): New function.
28377         * tree.h: Declared here.
28378         * tsan.c: Use sanitize_flags_p.
28379         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28380         (instrument_mem_ref): Likewise.
28381         (instrument_bool_enum_load): Likewise.
28382         (do_ubsan_in_current_function): Remove the function.
28383         (pass_ubsan::execute): Use sanitize_flags_p.
28384         * ubsan.h: Remove do_ubsan_in_current_function
28385         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28386         (dump_function_to_file): Use it here.
28388 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28390         PR tree-optimization/80803
28391         PR tree-optimization/81063
28392         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28393         (propagate_subaccesses_across_link): Enqueue subtree whenever
28394         necessary instead of relying on the caller.
28396 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28398         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28399         that have a first_link.
28400         (sort_and_splice_var_accesses): Do not check first_link before
28401         enquing.
28402         (subtree_mark_written_and_enqueue): Likewise.
28403         (propagate_all_subaccesses): Likewise and do not stop at first
28404         parent with a first_link.
28406 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28408         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28409         instead of f.
28411 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28413         * match.pd: New pattern.
28415 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28417         * tree-vrp.c (is_masked_range_test): New function.
28418         (register_edge_assert_for): Determine ranges for
28419         some bit tests.
28421 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28423         PR tree-optimization/67328
28424         * fold-const.c (maskable_range_p): New function.
28425         (build_range_check): Generate bittests if possible.
28427 2017-06-13  Martin Liska  <mliska@suse.cz>
28429         * gimple-pretty-print.c (dump_probability): Add new argument.
28430         (dump_edge_probability): Dump both probability and count.
28431         (dump_gimple_label): Likewise.
28432         (dump_gimple_bb_header): Likewise.
28434 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28436         PR target/81072
28437         * config/avr/avr-devices.c: Fix indentation.
28438         * config/avr/gen-avr-mmcu-specs.c: Dito.
28440 2017-06-13  Richard Biener  <rguenther@suse.de>
28442         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28443         instead get vector type from stmt_info.
28444         (vectorizable_reduction): Adjust.  Remove dead code.
28446 2017-06-13  Richard Biener  <rguenther@suse.de>
28448         PR middle-end/81065
28449         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28450         case of C * (x * C2 + C3).
28451         (fold_addr_of_array_ref_difference): Properly fold index difference.
28453 2017-06-12  David S. Miller  <davem@davemloft.net>
28455         PR target/80968
28456         * config/sparc/sparc.md (return expander): Emit frame blockage if
28457         function uses alloca.
28459 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28461         * combine.c (make_field_assignment): Check len rather than the mode
28462         precision when calling force_to_mode.
28464 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28466         Support multilibs and devices that see flash in RAM address range.
28468         PR target/81072
28469         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28470         (avr_mcu_t) <flash_pm_offset>: New field.
28471         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28472         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28473         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28474         (AVR_TINY_PM_OFFSET): Remove macro.
28475         * config/avr/avr.opt (-mshort-calls): New option.
28476         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28477         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28478         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28479         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28480         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28481         instead of avr_arch->have_jmp_call.
28482         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28483         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28484         avr_arch->flash_pm_offset to define.
28485         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28486         new field flash_pm_offset.  Add entry for avrxmega3.
28487         (avr_texinfo): Add entry for avrxmega3.
28488         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28489         attiny212, attiny214,
28490         attiny412, attiny414, attiny416, attiny417,
28491         attiny814, attiny816, attiny817,
28492         attiny1614, attiny1616, attiny1617,
28493         attiny3214, attiny3216, attiny3217.
28494         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28495         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28496         (avr_print_operand_address) [AVR_TINY]: Same.
28497         (avr_asm_init_sections) <readonly_data_section>: Only patch
28498         callback if avr_arch->flash_pm_offset = 0.
28499         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28500         for rodata if avr_arch->flash_pm_offset != 0.
28501         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28502         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28503         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28504         (m_options): Append opt_rcall.
28505         (m_dirnames): Append dir_rcall.
28506         * config/avr/t-multilib: Regenerate.
28508         * configure.ac [target=avr]: Check whether avrxmega3 default
28509         linker description file works as needed.
28510         * configure: Regenerate.
28511         * doc/avr-mmcu.texi: Regenerate.
28512         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28513         <__AVR_ARCH__>: Document avrxmega3 and 103.
28514         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28515         <__AVR_SHORT_CALLS__>: Document it.
28516         <__AVR_PM_BASE_ADDRESS__>: Document it.
28517         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28518         (AVR Variable Attributes) <progmem>: Document this is
28519         not needed for avrxmega3.
28520         (AVR Named Address Spaces) <__flash>: Dito.
28522 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28524         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28526 2017-06-12  Doug Rupp  <rupp@adacore.com>
28528         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28529         Append vxworks-stdint.h to the tm_file list.
28530         * config/vxworks-stdint.h: New file.
28532 2017-06-12  Martin Liska  <mliska@suse.cz>
28534         PR tree-optimization/81041
28535         * tree-profile.c (gimple_gen_ic_func_profiler):
28536         Create an extra BB in profile-generate
28537         (gimple_gen_time_profiler): Likewise.
28539 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28541         PR tree-optimization/81003
28542         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28543         (update_range_test): Use it instead of force_gimple_operand_gsi.
28545 2017-06-12  Richard Biener  <rguenther@suse.de>
28547         PR tree-optimization/81053
28548         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28549         with backedge value not defined in loop.  Simplify def stmt
28550         compute.
28552 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28554         PR target/79939
28555         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28556         Return true.
28557         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28558         nvptx_cannot_force_const_mem.
28560 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28562         * opts.c (finish_options): Move test for flag_split_stack after
28563         it has been initialized.
28565 2017-06-11  Jason Merrill  <jason@redhat.com>
28567         * tree.h (id_equal): New.
28568         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28569         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28570         instead of strcmp of IDENTIFIER_POINTER.
28572 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28574         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28575         (mark_all_inlined_calls_cdtor): Fix formating.
28576         (inline_transform): Rescale profile before inlining.
28578 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28580         * cgraph.h (cgraph_edge::clone): Update prototype.
28581         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28582         (cgraph_node::create_clone): Update.
28583         (cgraph_node::create_version_clone): Update.
28584         * tree-inline.c (copy_bb): Update.
28585         (expand_call_inline): Update.
28587 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28589         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28590         factored out from ...
28591         (rs6000_emit_prologue): ... here.
28593 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28595         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28596         factored out from ...
28597         (rs6000_emit_prologue): ... here.
28599 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28601         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28602         edge counts.
28603         (handle_missing_profiles): Fix computation of tp_first_run.
28604         (counts_to_freqs): Do not touch freqs when count is 0.
28606 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28608         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28609         profile.
28611 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28613         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28614         attributes): Document signal effective target.
28616 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28618         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28619         Document effective target stack_size.
28621 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28623         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28624         to the edit_context if they can be auto-applied.
28626 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28628         * opts.c (finish_options): If -fsplit-stack, disable implicit
28629         -forder-blocks-and-partition.
28630         * doc/invoke.texi (Optimize Options): Document that when using
28631         -fsplit-stack -forder-blocks-and-partition is not implicitly
28632         enabled.
28634 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28636         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28637         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28638         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28639         * builtins.def (abort, trap, unreachable): Declare cold.
28640         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28641         * tree-core.h (ECF_COLD): New.
28642         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28643         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28645 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28647         * predict.c (unlikely_executed_stmt_p): Cleanup.
28649 2017-06-09  Richard Biener  <rguenther@suse.de>
28651         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28652         model if the ref is always written to.
28654 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28656         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28658 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28660         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28661         than udiv.
28663 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28665         PR target/80855
28666         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28667         "target cannot support label values" when encountering LABEL_REF.
28669 2017-06-09  Martin Liska  <mliska@suse.cz>
28671         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28672         (gimple_gen_ic_func_profiler): Emit direct comparison
28673         of __gcov_indirect_call_callee with NULL.
28674         (gimple_gen_time_profiler): Change probability from
28675         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28677 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28679         * profile.c (edge_gcov_counts): Turn to pointer.
28680         (compute_branch_probabilities, compute_branch_probabilities): Update.
28681         (branch_prob): Do not clear edge_gcov_count.
28682         * profile.h (edge_gcov_counts): Turn to pointer.
28683         (edge_gcov_count): Update.
28685 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28687         * gimple.h (gimple_check_failed): Mark cold.
28689 2017-06-09  Richard Biener  <rguenther@suse.de>
28691         PR tree-optimization/66623
28692         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28693         refactor check_reduction into two parts, properly computing
28694         whether we have to check reduction validity for outer loop
28695         vectorization.
28697 2017-06-09  Richard Biener  <rguenther@suse.de>
28699         PR tree-optimization/79483
28700         * graphite-scop-detection.c (order): New global.
28701         (get_order): Compute bb to order mapping that satisfies code
28702         generation constraints.
28703         (cmp_pbbs): New helper.
28704         (build_scops): Start domwalk at entry block, sort generated
28705         pbbs.
28707 2017-06-09  Richard Biener  <rguenther@suse.de>
28709         PR middle-end/81007
28710         * ipa-polymorphic-call.c
28711         (ipa_polymorphic_call_context::restrict_to_inner_class):
28712         Skip FIELD_DECLs with error_mark_node type.
28713         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28714         last again.
28716 2017-06-09  Martin Liska  <mliska@suse.cz>
28718         * predict.c (struct branch_predictor): New struct.
28719         (test_prediction_value_range): New test.
28720         (predict_c_tests): New function.
28721         * selftest-run-tests.c (selftest::run_tests): Run the function.
28722         * selftest.h: Declare new tests.
28724 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28726         PR target/80966
28727         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28728         gen_add3_insn did not fail.
28729         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28730         r0, construct that number in a temporary reg and add that reg to r0.
28731         If asked to put the result in r0 as well, fail.
28733 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28735         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28736         for early expansion of vec_eqv.
28738 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28740         PR middle-end/81005
28741         * ubsan.c (instrument_null): Avoid pointless code temporary.
28742         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28744 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28746         PR target/81015
28747         Revert:
28748         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28750         PR target/59874
28751         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28752         (*clzhi2): Ditto.
28754 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28756         * predict.c (unlikely_executed_edge_p): Move ahead.
28757         (probably_never_executed_edge_p): Use it.
28759 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28761         PR middle-end/79988
28762         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28763         gimple_call_builtin_p call.
28765 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28767         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28768         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28769         rtl_check_failed_type2, rtl_check_failed_code1,
28770         rtl_check_failed_code2, rtl_check_failed_code_mode,
28771         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28772         rtvec_check_failed_bounds, rtl_check_failed_flag,
28773         _fatal_insn_not_found, _fatal_insn): Likewise.
28774         * tree.h (tree_contains_struct_check_failed,
28775         tree_check_failed, tree_not_check_failed,
28776         tree_class_check_failed, tree_range_check_failed,
28777         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28778         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28779         tree_operand_check_failed, omp_clause_check_failed,
28780         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28781         Likewise.
28783 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28785         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28786         flag_branch_probabilities.
28787         * ipa-inline.c (edge_badness): Likewise.
28788         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28789         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28790         * predict.c (maybe_hot_frequency_p): Likewise.
28791         (probably_never_executed): Likewise.
28792         * sched-ebb.c (schedule_ebbs): Likewise.
28793         * sched-rgn.c (find_single_block_region): Likewise.
28794         * tracer.c (tail_duplicate): Likewise.
28796 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28798         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28799         longer requires x_flag_profile_use.
28801 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28803         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28804         instead of flag_reorder_blocks_and_partition.
28805         * dbxout.c (dbxout_function_end): Likewise.
28806         * dwarf2out.c (gen_subprogram_die): Likewise.
28807         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28808         * hw-doloop.c (reorg_loops): Likewise.
28809         * varasm.c (assemble_start_function,
28810         assemble_end_function): Likewise.
28811         (decide_function_section): Do not check for
28812         flag_reorder_blocks_and_partition.
28814 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28816         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28817         New function.
28818         (chkp_get_hard_register_fake_addr_expr): Ditto.
28819         (chkp_build_addr_expr): Add check for hard reg case.
28820         (chkp_parse_array_and_component_ref): Ditto.
28821         (chkp_find_bounds_1): Ditto.
28822         (chkp_process_stmt): Don't generate bounds store for
28823         hard reg case.
28825 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28827         * predict.c (maybe_hot_bb_p): Do not check profile status.
28828         (maybe_hot_edge_p): Likewise.
28829         (probably_never_executed): Check for zero counts even if profile
28830         is not read.
28831         (unlikely_executed_edge_p): New function.
28832         (unlikely_executed_stmt_p): New function.
28833         (unlikely_executed_bb_p): New function.
28834         (set_even_probabilities): Use unlikely predicates.
28835         (combine_predictions_for_bb): Likewise.
28836         (predict_paths_for_bb): Likewise.
28837         (predict_paths_leading_to_edge): Likewise.
28838         (determine_unlikely_bbs): New function.
28839         (estimate_bb_frequencies): Use it.
28840         (compute_function_frequency): Use zero counts even if profile is
28841         not read.
28842         * profile-count.h: Fix typo.
28844 2017-08-08  Julia Koval  <julia.koval@intel.com>
28846         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28847         _mm512_mask_cvtsepi16_storeu_epi8,
28848         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28849         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28850         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28851         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28852         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28853         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28854         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28855         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28856         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28857         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28858         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28859         __builtin_ia32_pmovuswb256mem_mask,
28860         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28861         __builtin_ia32_pmovwb512mem_mask): New builtins.
28863 2017-08-08  Julia Koval  <julia.koval@intel.com>
28865         PR target/73350,80862
28866         * config/i386/subst.md (round): Fix round pattern.
28867         * config/i386/i386.c (ix86_erase_embedded_rounding):
28868         Fix erasing rounding for the fixed pattern.
28870 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28872         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
28874 2017-06-08  Martin Liska  <mliska@suse.cz>
28876         PR gcov-profile/80911
28877         * gcov.c (block_info::block_info): New constructor.
28879 2017-06-07  Carl Love  <cel@us.ibm.com>
28881         * config/rs6000/rs6000-c: The return type of the following
28882         built-in functions was implemented as int not long long.  Fix sign
28883         of return value for the unsigned version of vec_mulo and vec_mule.
28884         vector unsigned long long vec_bperm (vector unsigned long long,
28885                                              vector unsigned char)
28886         vector signed long long vec_mule (vector signed int,
28887                                           vector signed int)
28888         vector unsigned long long vec_mule (vector unsigned int,
28889                                             vector unsigned int)
28890         vector signed long long vec_mulo (vector signed int,
28891                                           vector signed int)
28892         vector unsigned long long vec_mulo (vector unsigned int,
28893                                             vector unsigned int)
28894         * doc/extend.texi: Fix the documentation for the built-in
28895         functions.
28897 2017-06-07  Carl Love  <cel@us.ibm.com>
28899         PR target/80982
28900         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
28901         for BE.
28903 2017-06-07  Carl Love  <cel@us.ibm.com>
28905         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
28906         support, Generate       doublehv for signed int/float for BE case only.
28908 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
28910         * doc/invoke.texi (mcx16): Rewrite.
28912 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28914         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
28915         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
28916         *mov<mode>_softfloat, and an anonymous splitter): Use
28917         nonimmediate_operand instead of rs6000_nonimmediate_operand.
28919 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28921         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
28922         SPEFSCR registers.
28923         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28924         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
28925         (rs6000_debug_reg_global): Adjust.
28926         (rs6000_init_hard_regno_mode_ok): Adjust.
28927         (rs6000_dbx_register_number): Adjust.
28928         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
28929         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28930         Remove SPE_ACC and SPEFSCR.
28931         (REG_ALLOC_ORDER): Ditto.
28932         (FRAME_POINTER_REGNUM): Change to 111.
28933         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
28934         (REG_CLASS_NAMES): Ditto.
28935         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
28936         (REGISTER_NAMES): Ditto.
28937         (ADDITIONAL_REG_NAMES): Ditto.
28938         (rs6000_reg_names): Ditto.
28939         * config/rs6000/rs6000.md: Renumber some register number
28940         define_constants.
28942 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28944         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
28945         registers.
28946         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28947         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
28948         to 117.
28949         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
28950         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28951         Delete the SPE high registers.
28952         (REG_ALLOC_ORDER): Ditto.
28953         (enum reg_class): Remove SPE_HIGH_REGS.
28954         (REG_CLASS_NAMES): Ditto.
28955         (REG_CLASS_CONTENTS): Delete the SPE high registers.
28956         (REGISTER_NAMES): Ditto.
28957         (rs6000_reg_names): Ditto.
28958         * doc/tm.texi.in: Remove SPE as example.
28959         * doc/tm.texi: Regenerate.
28961 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28963         * config/rs6000/8540.md (ppc8540_brinc): Delete.
28964         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
28965         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
28966         * config/rs6000/rs6000.md (type): Remove "brinc".
28968 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28970         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
28971         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
28972         * config/rs6000/linuxspe.h: Delete file.
28973         * config/rs6000/rs6000.md: Don't include spe.md.
28974         * config/rs6000/spe.h: Delete file.
28975         * config/rs6000/spe.md: Delete file.
28976         * config/rs6000/t-rs6000: Remove spe.md.
28978 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28980         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
28981         (reg_or_none500mem_operand): Delete.
28982         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
28983         instead of reg_or_none500mem_operand.
28985 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28987         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
28988         handling of SPE flags.
28989         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
28991 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28993         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
28994         SPE ABI handling.
28995         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
28996         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
28997         paired_divv2sf3): Similar.
28998         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
28999         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29000         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29001         RS6000_BUILTIN_S.
29002         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29003         Rename the paired_* instruction patterns.
29004         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29005         define __SPE__.
29006         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29007         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29008         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29009         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29010         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29011         PAIRED_VECTOR_MODE.
29012         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29013         (spe_func_has_64bit_regs_p): Delete.
29014         (spe_expand_predicate_builtin): Delete.
29015         (spe_expand_evsel_builtin): Delete.
29016         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29017         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29018         (invalid_e500_subreg): Delete.
29019         (rs6000_legitimize_address): Always force_reg op2 as well, for
29020         paired single memory accesses.
29021         (rs6000_member_type_forces_blk): Delete.
29022         (rs6000_spe_function_arg): Delete.
29023         (rs6000_expand_unop_builtin): Delete SPE handling.
29024         (rs6000_expand_binop_builtin): Ditto.
29025         (spe_expand_stv_builtin): Delete.
29026         (bdesc_2arg_spe): Delete.
29027         (spe_expand_builtin): Delete.
29028         (spe_expand_predicate_builtin): Delete.
29029         (spe_expand_evsel_builtin): Delete.
29030         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29031         (spe_init_builtins): Delete.
29032         (spe_func_has_64bit_regs_p): Delete.
29033         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29034         (rs6000_emit_stack_reset): Ditto.
29035         (rs6000_dwarf_register_span): Delete.
29036         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29037         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29038         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29039         Delete.
29040         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29041         Delete.
29042         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29043         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29044         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29045         mulv2sf3, divv2sf3): Delete expanders.
29047 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29049         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29051 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29053         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29054         * config/rs6000/rs6000.c: Ditto.
29056 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29058         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29059         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29060         comparison_operator.
29062 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29064         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29065         * config/rs6000/rs6000.opt: Ditto.
29066         * config/rs6000/t-rtems: Ditto.
29068 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29070         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29071         TARGET_E500_SINGLE by 0, simplify.
29072         * config/rs6000/rs6000.c: Ditto.
29073         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29074         (spe_build_register_parallel): Delete.
29075         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29076         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29077         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29078         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29079         (E500_CONVERT): Delete.
29080         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29082 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29084         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29085         * config/rs6000/dfp.md: Ditto.
29086         (negdd2, *negdd2_fpr): Merge.
29087         (absdd2, *absdd2_fpr): Merge.
29088         (negtd2, *negtd2_fpr): Merge.
29089         (abstd2, *abstd2_fpr): Merge.
29090         * config/rs6000/e500.h: Delete file.
29091         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29092         TARGET_FPRS by 1 and simplify.
29093         * config/rs6000/rs6000-c.c: Ditto.
29094         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29095         TARGET_DF_SPE by 0.
29096         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29097         TARGET_DF_SPE.
29098         * config/rs6000/rs6000.md: Ditto.
29099         (floatdidf2, *floatdidf2_fpr): Merge.
29100         (move_from_CR_gt_bit): Delete.
29101         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29102         (E500_CR_IOR_COMPARE): Delete.
29103         (All patterns that require !TARGET_FPRS): Delete.
29104         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29106 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29108         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29110 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29112         * graphds.c (add_edge): Intitialize edge's attached data.
29113         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29114         pointer parameter.  Call pointed function on each edge during
29115         graph traversing.  Skip traversing the edge when the function
29116         returns true.
29117         (graphds_dfs, graphds_scc): Ditto.
29118         (for_each_edge): New parameter.  Pass the new parameter to callback
29119         function.
29120         * graphds.h (skip_edge_callback): New function pointer type.
29121         (graphds_dfs, graphds_scc): New function pointer parameter.
29122         (graphds_edge_callback, for_each_edge): New parameter.
29124 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29126         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29127         out code checking if runtime alias check is possible to below ...
29128         Call the new function.
29129         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29130         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29132 2017-06-07  Marek Polacek  <polacek@redhat.com>
29134         PR sanitizer/80932
29135         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29136         TYPE_OVERFLOW_WRAPS check.
29138 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29140         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29141         if versioning is required.
29142         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29143         peeling with the check for versioning.
29145 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29147         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29148         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29149         Set true to new parameter if new ssa variable is defined.
29150         (vect_gen_vector_loop_niters): Refactor.  Set range information
29151         for the new vector loop bound variable.
29152         (vect_do_peeling): Ditto.
29154 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29156         * tree-affine.c (ssa.h): Include header file.
29157         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29158         has wrapping overflow behavior.
29160 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29162         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29164 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29166         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29167         (tree_to_aff_combination): ... here.
29169 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29171         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29172         reg_pressure model function.
29173         (ivopts_global_cost_for_size): Delete.
29174         (determine_set_costs, iv_ca_recount_cost): Call new model function
29175         ivopts_estimate_reg_pressure.
29177 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29179         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29180         expensive than udiv.  Remove floating point cases from mod.
29182 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29184         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29185         Increase idiv cost.
29187 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29189         * config/aarch64/aarch64.md
29190         (copysignsf3): Fix mask generation.
29192 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29194         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29195         TDI_gimple.
29196         (class dump_manager): Add register_dumps method.
29197         * dumpfile.c: Include langhooks.h.
29198         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29199         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29200         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29201         (dump_manager::dump_register): Start with 512 entries instead of 32.
29202         (dump_manager::register_dumps): New method.
29203         * toplev.c (general_init): Instead of invoking register_dumps
29204         langhook, invoke register_dumps method on the dump manager.
29205         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29206         TDI_generic.
29208 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29210         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29211         Say that # requires an associated define_split to exist, and that
29212         the define_split must be suitable for use after register allocation.
29214 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29216         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29217         (compute_outgoing_frequencies): Also initialize zero counts.
29218         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29219         around loops; preserve more of profile when nothing changes.
29221 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29223         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29224         here.
29225         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29226         * config/arm/arm-cpu-cdata.h: Regenerate.
29227         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29228         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29229         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29230         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29231         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29232         support.
29233         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29234         support.
29235         * config/arm/t-rmprofile: Likewise.
29236         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29238 2017-06-06  David S. Miller  <davem@davemloft.net>
29240         PR target/80968
29241         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29242         blockage if function uses alloca.
29244 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29246         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29247         New "uid" fields to hold pretty-print IDs of group and ref.
29248         Memory references are now identified as <group_id>:<ref_id>
29249         instead of using [random] addresses.
29250         (dump_mem_details): Simplify, no functional change.
29251         (dump_mem_ref): Simplify and make output more concise.
29252         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29253         (find_or_create_group): Initialize group uid.
29254         (record_ref): Initialize ref uid.  Improve debug output.
29255         (prune_group_by_reuse, should_issue_prefetch_p,)
29256         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29257         (mark_nontemporal_store, determine_loop_nest_reuse):
29258         Improve debug output.
29260 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29262         * dbgcnt.def (prefetch): New debug counter.
29263         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29264         (schedule_prefetches): Stop issueing prefetches if debug counter
29265         tripped.
29267 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29269         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29270         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29272 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29274         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29275         Use aarch64_reg_or_zero predicate for operand 4.
29276         (aarch64_compare_and_swap<mode> define_insn_and_split):
29277         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29278         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29280 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29282         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29283         (arm_compute_save_core_reg_mask): This.
29284         (thumb1_compute_save_reg_mask): Rename into ...
29285         (thumb1_compute_save_core_reg_mask): This.
29286         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29287         (arm_compute_frame_layout): Likewise.
29289 2017-06-06  Richard Biener  <rguenther@suse.de>
29291         PR tree-optimization/80974
29292         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29293         keep or clear leaders SSA info.
29295 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29297         * config/nvptx/nvptx.c (split_mode_p): New function.
29298         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29300 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29302         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29304 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29306         PR bootstrap/80978
29307         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29308         profile.
29310 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29312         * shrink-wrap.c (handle_simple_exit): Update profile.
29313         (try_shrink_wrapping): Upate profile.
29315 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29317         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29318         (tree_guess_outgoing_edge_probabilities): New.
29319         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29320         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29322 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29324         * ipa-split.c (split_function): Initialize return bb profile.
29326 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29328         * profile.c (compute_branch_probabilities): Also initialize
29329         EXIT_BLOCK profile.
29331 2017-06-06  Richard Biener  <rguenther@suse.de>
29333         PR tree-optimization/80928
29334         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29335         (vect_analyze_loop_operations): Properly guard analysis for
29336         pure SLP case.
29337         (vect_transform_loop): Likewise.
29338         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29339         (vect_model_induction_cost): Do not cost for pure SLP.
29340         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29341         of induction in inner loop vectorization.
29342         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29343         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29344         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29345         recursion.
29346         (vect_analyze_slp_cost_1): Cost induction.
29347         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29348         (vect_get_slp_vect_defs): Likewise.
29349         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29350         (vect_transform_stmt): Handle SLP reductions.
29351         * tree-vectorizer.h (vectorizable_induction): Adjust.
29353 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29355         * config/rs6000/rs6000.c (make_resolver_func): Update
29356         init_lowered_empty_function call.
29358 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29360         * doc/invoke.texi: Document the -fprofile-abs-path option.
29361         * common.opt (fprofile-abs-path): New option.
29362         * gcov-io.h (gcov_write_filename): Declare.
29363         * gcov-io.c (gcov_write_filename): New function.
29364         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29365         * profile.c (output_location): Likewise.
29367 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29369         * shring-wrap.c: Revert accidental commit.
29371 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29373         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29375 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29377         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29378         new edge.
29379         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29380         profile in callgraph edge.
29381         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29382         (apply_scale): Likewise.
29383         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29384         Also scale profile when inlining function with zero profile.
29385         (initialize_cfun): Update exit block profile even when it is zero.
29386         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29387         when profile is read.
29389 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29391         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29392         (CLONE_*): New constants to define the processors we can generate
29393         code for with the target_clone attribute.
29394         (rs6000_clone_map): New array to identify which clone processors
29395         the current program is running on.
29396         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29397         target_clone attribute.
29398         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29399         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29400         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29401         (cpu_expand_builtin): Add support for target_clone attribute.
29402         (rs6000_valid_attribute_p): Allow "default" attribute.
29403         (get_decl_name): New debug function to simplify printing the
29404         current function name in debugging statements.
29405         (rs6000_clone_priority): New functions to support the target_clone
29406         attribute, and be able to generate code to switch between ISA 2.05
29407         through ISA 3.0 (power6 through power9).
29408         (rs6000_compare_version_priority): Likewise.
29409         (rs6000_get_function_versions_dispatcher): Likewise.
29410         (make_resolver_func): Likewise.
29411         (add_condition_to_bb): Likewise.
29412         (dispatch_function_versions): Likewise.
29413         (rs6000_generate_version_dispatcher_body): Likewise.
29414         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29415         (fusion_gpr_load_p): Fix a spacing issue.
29416         * doc/extend.texi (Common Function Attributes): Document that the
29417         PowerPC supports the target_clone attribute.
29419 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29421         * config/arm/arm.h: explain F symbol found in description of ARM
29422         register allocation in its legend.
29424 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29426         * config/mips/frame-header-opt.c: Include profile-count.h.
29427         * config/riscv/riscv.c: Include profile-count.h
29429 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29431         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29432         update profile.
29433         (sm_set_flag_if_changed): Add bbs field.
29434         (execute_sm_if_changed_flag_set): Pass BBS.
29435         (execute_sm): Update.
29437 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29439         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29440         New pattern.
29442 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29444         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29445         (peephole2): New peephole2 to emit the above.
29446         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29448 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29450         * config/aarch64/aarch64.c (define_peephole2 above
29451         *sub_<shift>_<mode>): New peephole.
29453 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29455         * config/i386/i386.c (make_resolver_func): Update.
29456         * Makefile.in: Add profile-count.h and profile-count.o
29457         * auto-profile.c (afdo_indirect_call): Update to new API.
29458         (afdo_set_bb_count): Update.
29459         (afdo_propagate_edge): Update.
29460         (afdo_propagate_circuit): Update.
29461         (afdo_calculate_branch_prob): Update.
29462         (afdo_annotate_cfg): Update.
29463         * basic-block.h: Include profile-count.h
29464         (struct edge_def): Turn count to profile_count.
29465         (struct basic_block_def): Likewie.
29466         (REG_BR_PROB_BASE): Move to profile-count.h
29467         (RDIV): Move to profile-count.h
29468         * bb-reorder.c (max_entry_count): Turn to profile_count.
29469         (find_traces): Update.
29470         (rotate_loop):Update.
29471         (connect_traces):Update.
29472         (sanitize_hot_paths):Update.
29473         * bt-load.c (migrate_btr_defs): Update.
29474         * cfg.c (RDIV): Remove.
29475         (init_flow): Use alloc_block.
29476         (alloc_block): Uninitialize count.
29477         (unchecked_make_edge): Uninitialize count.
29478         (check_bb_profile): Update.
29479         (dump_edge_info): Update.
29480         (dump_bb_info): Update.
29481         (update_bb_profile_for_threading): Update.
29482         (scale_bbs_frequencies_int): Update.
29483         (scale_bbs_frequencies_gcov_type): Update.
29484         (scale_bbs_frequencies_profile_count): New.
29485         * cfg.h (update_bb_profile_for_threading): Update.
29486         (scale_bbs_frequencies_profile_count): Declare.
29487         * cfgbuild.c (compute_outgoing_frequencies): Update.
29488         (find_many_sub_basic_blocks): Update.
29489         * cfgcleanup.c (try_forward_edges): Update.
29490         (try_crossjump_to_edge): Update.
29491         * cfgexpand.c (expand_gimple_tailcall): Update.
29492         (construct_exit_block): Update.
29493         * cfghooks.c (verify_flow_info): Update.
29494         (dump_bb_for_graph): Update.
29495         (split_edge): Update.
29496         (make_forwarder_block): Update.
29497         (duplicate_block): Update.
29498         (account_profile_record): Update.
29499         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29500         (get_estimated_loop_iterations): Update.
29501         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29502         (single_likely_exit): Update.
29503         * cfgloopmanip.c (scale_loop_profile): Update.
29504         (loopify): Update.
29505         (set_zero_probability): Update.
29506         (lv_adjust_loop_entry_edge): Update.
29507         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29508         (purge_dead_edges): Update.
29509         (rtl_account_profile_record): Update.
29510         * cgraph.c (cgraph_node::create): Uninitialize count.
29511         (symbol_table::create_edge): Uninitialize count.
29512         (cgraph_update_edges_for_call_stmt_node): Update.
29513         (cgraph_edge::dump_edge_flags): Update.
29514         (cgraph_node::dump): Update.
29515         (cgraph_edge::maybe_hot_p): Update.
29516         * cgraph.h: Include profile-count.h
29517         (create_clone), create_edge, create_indirect_edge): Update.
29518         (cgraph_node): Turn count to profile_count.
29519         (cgraph_edge0: Likewise.
29520         (make_speculative, clone): Update.
29521         (create_edge): Update.
29522         (init_lowered_empty_function): Update.
29523         * cgraphclones.c (cgraph_edge::clone): Update.
29524         (duplicate_thunk_for_node): Update.
29525         (cgraph_node::create_clone): Update.
29526         * cgraphunit.c (cgraph_node::analyze): Update.
29527         (cgraph_node::expand_thunk): Update.
29528         * final.c (dump_basic_block_info): Update.
29529         * gimple-streamer-in.c (input_bb): Update.
29530         * gimple-streamer-out.c (output_bb): Update.
29531         * graphite.c (print_global_statistics): Update.
29532         (print_graphite_scop_statistics): Update.
29533         * hsa-brig.c: Include basic-block.h.
29534         * hsa-dump.c: Include basic-block.h.
29535         * hsa-gen.c (T sum_slice): Update.
29536         (convert_switch_statements):Update.
29537         * hsa-regalloc.c: Include basic-block.h.
29538         * ipa-chkp.c (chkp_produce_thunks): Update.
29539         * ipa-cp.c (struct caller_statistics): Update.
29540         (init_caller_stats): Update.
29541         (gather_caller_stats): Update.
29542         (ipcp_cloning_candidate_p): Update.
29543         (good_cloning_opportunity_p): Update.
29544         (get_info_about_necessary_edges): Update.
29545         (dump_profile_updates): Update.
29546         (update_profiling_info): Update.
29547         (update_specialized_profile): Update.
29548         (perhaps_add_new_callers): Update.
29549         (decide_about_value): Update.
29550         (ipa_cp_c_finalize): Update.
29551         * ipa-devirt.c (struct odr_type_warn_count): Update.
29552         (struct decl_warn_count): Update.
29553         (struct final_warning_record): Update.
29554         (possible_polymorphic_call_targets): Update.
29555         (ipa_devirt): Update.
29556         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29557         * ipa-icf.c (sem_function::merge): Update.
29558         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29559         * ipa-inline.c (compute_uninlined_call_time): Update.
29560         (compute_inlined_call_time): Update.
29561         (want_inline_small_function_p): Update.
29562         (want_inline_self_recursive_call_p): Update.
29563         (edge_badness): Update.
29564         (lookup_recursive_calls): Update.
29565         (recursive_inlining): Update.
29566         (inline_small_functions): Update.
29567         (dump_overall_stats): Update.
29568         (dump_inline_stats): Update.
29569         * ipa-profile.c (ipa_profile_generate_summary): Update.
29570         (ipa_propagate_frequency): Update.
29571         (ipa_profile): Update.
29572         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29573         * ipa-utils.c (ipa_merge_profiles): Update.
29574         * loop-doloop.c (doloop_modify): Update.
29575         * loop-unroll.c (report_unroll): Update.
29576         (unroll_loop_runtime_iterations): Update.
29577         * lto-cgraph.c (lto_output_edge): Update.
29578         (lto_output_node): Update.
29579         (input_node): Update.
29580         (input_edge): Update.
29581         (merge_profile_summaries): Update.
29582         * lto-streamer-in.c (input_cfg): Update.
29583         * lto-streamer-out.c (output_cfg): Update.
29584         * mcf.c (create_fixup_graph): Update.
29585         (adjust_cfg_counts): Update.
29586         (sum_edge_counts): Update.
29587         * modulo-sched.c (sms_schedule): Update.
29588         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29589         * predict.c (maybe_hot_count_p): Update.
29590         (probably_never_executed): Update.
29591         (dump_prediction): Update.
29592         (combine_predictions_for_bb): Update.
29593         (propagate_freq): Update.
29594         (handle_missing_profiles): Update.
29595         (counts_to_freqs): Update.
29596         (rebuild_frequencies): Update.
29597         (force_edge_cold): Update.
29598         * predict.h: Include profile-count.h
29599         (maybe_hot_count_p, counts_to_freqs): UPdate.
29600         * print-rtl-function.c: Do not include cfg.h
29601         * print-rtl.c: Include basic-block.h
29602         * profile-count.c: New file.
29603         * profile-count.h: New file.
29604         * profile.c (is_edge_inconsistent): Update.
29605         (correct_negative_edge_counts): Update.
29606         (is_inconsistent): Update.
29607         (set_bb_counts): Update.
29608         (read_profile_edge_counts): Update.
29609         (compute_frequency_overlap): Update.
29610         (compute_branch_probabilities): Update; Initialize and deinitialize
29611         gcov_count tables.
29612         (branch_prob): Update.
29613         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29614         (edge_gcov_count): New.
29615         (bb_gcov_count): New.
29616         * shrink-wrap.c (try_shrink_wrapping): Update.
29617         * tracer.c (better_p): Update.
29618         * trans-mem.c (expand_transaction): Update.
29619         (ipa_tm_insert_irr_call): Update.
29620         (ipa_tm_insert_gettmclone_call): Update.
29621         * tree-call-cdce.c: Update.
29622         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29623         (gimple_duplicate_sese_tail): Update.
29624         (gimple_account_profile_record): Update.
29625         (execute_fixup_cfg): Update.
29626         * tree-inline.c (copy_bb): Update.
29627         (copy_edges_for_bb): Update.
29628         (initialize_cfun): Update.
29629         (freqs_to_counts): Update.
29630         (copy_cfg_body): Update.
29631         (expand_call_inline): Update.
29632         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29633         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29634         (try_unroll_loop_completely): Update.
29635         (try_peel_loop): Update.
29636         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29637         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29638         * tree-ssa-loop-split.c (connect_loops): Update.
29639         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29640         * tree-ssa-reassoc.c (branch_fixup): Update.
29641         * tree-ssa-tail-merge.c (replace_block_by): Update.
29642         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29643         (compute_path_counts): Update.
29644         (update_profile): Update.
29645         (recompute_probabilities): Update.
29646         (update_joiner_offpath_counts): Update.
29647         (estimated_freqs_path): Update.
29648         (freqs_to_counts_path): Update.
29649         (clear_counts_path): Update.
29650         (ssa_fix_duplicate_block_edges): Update.
29651         (duplicate_thread_path): Update.
29652         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29653         (struct switch_conv_info): Update.
29654         * tree-tailcall.c (decrease_profile): Update.
29655         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29656         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29657         * value-prof.c (check_counter): Update.
29658         (gimple_divmod_fixed_value): Update.
29659         (gimple_mod_pow2): Update.
29660         (gimple_mod_subtract): Update.
29661         (gimple_ic_transform): Update.
29662         (gimple_stringop_fixed_value): Update.
29663         * value-prof.h (gimple_ic): Update.
29665 2017-06-02  Carl Love  <cel@us.ibm.com>
29667         * config/rs6000/rs6000-c: Add support for built-in functions
29668         vector double vec_doublee (vector signed int);
29669         vector double vec_doublee (vector unsigned int);
29670         vector double vec_doublee (vector float);
29671         vector double vec_doubleh (vector signed int);
29672         vector double vec_doubleh (vector unsigned int);
29673         vector double vec_doubleh (vector float);
29674         vector double vec_doublel (vector signed int);
29675         vector double vec_doublel (vector unsigned int);
29676         vector double vec_doublel (vector float);
29677         vector double vec_doubleo (vector signed int);
29678         vector double vec_doubleo (vector unsigned int);
29679         vector double vec_doubleo (vector float);.
29680         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29681         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29682         UNS_DOUBLEL.
29683         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29684         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29685         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29686         VS_sxwsp.
29687         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29688         vec_doublel, vec_doubleh.
29689         * doc/extend.texi: Update the built-in documentation file for the
29690         new built-in functions.
29692 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29694         PR jit/80954
29695         * ipa-inline-analysis.c (free_growth_caches): Set
29696         edge_removal_hook_holder to NULL after removing it.
29698 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29700         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29701         comparision with zero.
29703 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29704         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29705         for early expansion of vec_min and vec_max builtins.
29706         (builtin_function_type): Add min/max unsigned variants to those
29707         identified as having unsigned arguments.
29709 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29711         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29713 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29715         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29716         Use VALL_F16 iterator rather than VALL.
29718 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29720         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29721         Emit CBNZ inside loop when doing a strong exchange and comparing
29722         against zero.  Generate the CC flags after the loop.
29724 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29726         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29727         (dl_section_ref): New.
29728         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29729         On AIX, append an expression to subtract the size of the
29730         section length to dl_section_ref.
29732 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29734         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29735         for early expansion of vector absolute builtins.
29737 2017-06-02  Richard Biener  <rguenther@suse.de>
29739         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29740         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29742 2017-06-02  Richard Biener  <rguenther@suse.de>
29744         PR tree-optimization/80948
29745         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29746         stmt order as well.
29748 2017-06-02  Richard Biener  <rguenther@suse.de>
29750         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29751         PHIs are ok.
29752         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29753         for inductions as relevant.
29755 2017-06-02  Richard Biener  <rguenther@suse.de>
29757         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29758         (vectorizable_induction): ... this.  Remove dead code.
29760 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29762         * builtins. (expand_builtin_alloca): Remove second parameter and
29763         infer its value from the first parameter instead.
29764         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29766 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29768         PR rtl-optimization/80903
29769         * loop-doloop.c (add_test): Unshare sequence.
29771 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29773         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29775 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29777         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29778         static.
29779         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29780         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29781         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29782         (xlogue_layout::get_stub_rtx): Make static.
29783         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29784         (xlogue_layout::compute_stub_managed_regs): Rename to...
29785         (xlogue_layout::count_stub_managed_regs): ...this.
29786         (xlogue_layout::is_stub_managed_reg): New function.
29787         (xlogue_layout::m_stub_names): Rename to...
29788         (xlogue_layout::s_stub_names): ...this, make static.
29789         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29790         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29791         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29792         xlogue_layout::s_stub_names): Instantiate statics.
29793         (stub_managed_regs): Remove.
29794         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29795         (disable_call_ms2sysv_xlogues): Rename to...
29796         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29797         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29798         warning logic.
29799         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29800         change after reload_completed.
29801         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29802         directly.
29803         (ix86_expand_prologue): Likewise.
29804         (ix86_expand_epilogue): Likewise.
29805         (ix86_expand_split_stack_prologue): Likewise.
29806         (ix86_compute_frame_layout): Remove frame parameter ...
29807         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29808         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29809         only if necessary.
29810         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29811         (ix86_frame): Move from here ...
29812         * config/i386/i386.h (ix86_frame): ... to here.
29813         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29814         complete ix86_frame data structure instead.  Remove some_ld_name.
29816 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29818         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29819         symbols that hold a DECL_VALUE_EXPR.
29821 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29823         PR tree-optimization/80898
29824         * tree-sra.c (process_subtree_disqualification): Removed.
29825         (disqualify_candidate): Do not acll
29826         process_subtree_disqualification.
29827         (subtree_mark_written_and_enqueue): New function.
29828         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29829         RHS has been disqualified and re-queue LHS if necessary.  Apart
29830         from that, ignore disqualified RHS.
29832 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29834         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29835         address fetch for z10 or later.
29837 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29839         * config/arc/arc.md (tst_movb): Add guard when splitting.
29841 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29843         * config/arc/arc.c (arc_can_eliminate): Test against
29844         arc_frame_pointer_needed.
29846 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29848         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29849         to prevent store reordering.
29850         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29851         (type): Add block type.
29852         (stack_tie): Define special instruction to be used in
29853         expand_prologue.
29855 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29857         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29858         constraint. It is not valid for the pattern.
29859         (noncommutative_binary_comparison): Likewise.
29861 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29863         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
29864         scaled addresses.
29866 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29868         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
29869         be used by the reg-alloc.
29871 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29873         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
29874         reg-alloc when having mul64 or mul32x16 instructions.
29875         (mulsidi3): Likewise.
29876         (umulsidi3): Likewise.
29877         (mulsi32x16): New pattern.
29878         (mulsi64): Likewise.
29879         (mulsidi64): Likewise.
29880         (umulsidi64): Likewise.
29881         (MUL32x16_REG): Define.
29882         (mul64_600): Use MUL32x16_REG.
29883         (mac64_600): Likewise.
29884         (umul64_600): Likewise.
29885         (umac64_600): Likewise.
29887 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29889         * config/arc/arc.md (mulsi3_700): Make it commutative.
29891 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
29893         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
29894         type for movstouw.
29895         (*sign_extendsidi2_insn): Likewise for movstosw.
29897 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29899         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
29900         the type of the input discriminant value.  Convert the
29901         discriminant value of signedness vary.
29903 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
29905         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
29906         Add to -Wall section.
29908 2017-06-01  Richard Biener  <rguenther@suse.de>
29910         PR middle-end/66313
29911         * fold-const.c (fold_plusminus_mult_expr): If the factored
29912         factor may be zero use a wrapping type for the inner operation.
29913         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
29914         and handle moved defs.
29915         (process_assignment): Properly guard the unary op case.  Return a
29916         tri-state indicating that moving the stmt before the call may allow
29917         to continue.  Pass through to_move.
29918         (find_tail_calls): Handle moving unrelated defs before
29919         the call.
29921 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
29923         PR target/80618
29924         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
29925         splitter result in the canonical way.
29927 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
29929         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
29930         also for 32bit target.  Update insn attributes.
29931         (zero-extendsidi2 splitter): Allow all registers for operand 1.
29933 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
29935         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
29936         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
29937         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
29938         (_mm_maskz_min_ss): New intrinsics.
29940 2017-05-31  Martin Liska  <mliska@suse.cz>
29942         * tree-vect-loop.c (vect_create_epilog_for_reduction):
29943         Change comment style to one we normally use.
29944         (vectorizable_reduction): Likewise.
29945         (vectorizable_induction): Likewise.
29946         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
29947         (vectorizable_call): Likewise.
29948         (vectorizable_simd_clone_call): Likewise.
29949         (vectorizable_conversion): Likewise.
29950         (vectorizable_assignment): Likewise.
29951         (vectorizable_shift): Likewise.
29952         (vectorizable_operation): Likewise.
29953         (vectorizable_store): Likewise.
29954         (vectorizable_load): Likewise.
29955         * tree-vectorizer.h: Likewise.
29957 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
29959         * passes.c (emergency_dump_function): New.
29960         * tree-pass.h (emergency_dump_function): Declare.
29961         * plugin.c (plugins_internal_error_function): Remove.
29962         * plugin.h (plugins_internal_error_function): Remove declaration.
29963         * toplev.c (internal_error_function): New static function.  Use it...
29964         (general_init): ...here.
29966 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
29968         * config/arc/arc.c (arc_print_operand): Handle constant operands.
29969         (arc_rtx_costs): Add costs for new patterns.
29970         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
29971         * config/arc/predicates.md: Add _1_2_3_operand predicate.
29973 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
29975         * tree-ssa-strlen.c (get_next_strinfo): New function.
29976         (get_stridx_plus_constant): Use it.
29977         (zero_length_string): Likewise.
29978         (adjust_related_strinfos): Likewise.
29979         (adjust_last_stmt): Likewise.
29981 2017-05-31  Richard Biener  <rguenther@suse.de>
29983         PR target/80880
29984         * config/i386/i386.c (ix86_expand_builtin): Remove assert
29985         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
29987 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
29989         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
29990         loop_vinfo argument and use of dependence distance vectors.
29991         Check instead whether the two references differ only in their
29992         initial value and assume that they have the same alignment if the
29993         difference is a multiple of the vector alignment.
29994         (vect_analyze_data_refs_alignment): Update call accordingly.
29996 2017-05-31  Martin Liska  <mliska@suse.cz>
29998         PR target/79155
29999         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30001 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30003         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30004         (create_intersect_range_checks): Move from ...
30005         * tree-data-ref.c (create_intersect_range_checks_index)
30006         (create_intersect_range_checks): ... to here.
30007         (create_runtime_alias_checks): New function factored from ...
30008         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30009         here.  Call above function.
30010         * tree-data-ref.h (create_runtime_alias_checks): New function.
30012 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30014         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30015         segment length for dr_b and compute it in wide_int.
30017 2017-05-31  Richard Biener  <rguenther@suse.de>
30019         PR tree-optimization/80906
30020         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30021         and pass through iv_map.
30022         (copy_bb_and_scalar_dependences): Adjust.
30023         (translate_pending_phi_nodes): Likewise.
30024         (copy_loop_close_phi_args): Handle code-generating IVs instead
30025         of ICEing.
30027 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30029         * diagnostic-color.c (color_dict): Add "type-diff".
30030         (parse_gcc_colors): Update comment.
30031         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30032         -fdiagnostics-show-template-tree and -fno-elide-type.
30033         (GCC_COLORS): Add type-diff to example.
30034         (type-diff=): New.
30035         (-fdiagnostics-show-template-tree): New.
30036         (-fno-elide-type): New.
30037         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30038         the pp_format_decoder callback.  Call any m_format_postprocessor's
30039         "handle" method.
30040         (pretty_printer::pretty_printer): Initialize
30041         m_format_postprocessor.
30042         (pretty_printer::~pretty_printer): Delete any
30043         m_format_postprocessor.
30044         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30045         (class format_postprocessor): New class.
30046         (struct pretty_printer::format_decoder): Document the new parameters.
30047         (struct pretty_printer::m_format_postprocessor): New field.
30048         * tree-diagnostic.c (default_tree_printer): Update for new
30049         bool and const char ** params.
30050         * tree-diagnostic.h (default_tree_printer): Likewise.
30052 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30054         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30055         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30056         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30057         rs6000_gen_cell_microcode code.
30058         (rs6000_final_prescan_insn): Delete.
30059         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30060         "warn-cell-microcode" entries.
30061         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30062         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30063         throughout.  Change cc_reg_not_micro_cr0_operand to
30064         cc_reg_not_cr0_operand throughout.
30065         (*extendhi<mode>2_noload): Delete.
30066         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30067         (mwarn-cell-microcode): Delete.
30068         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30069         -mgen-cell-microcode and -mwarn-cell-microcode.
30071 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30073         PR target/80833
30074         * config/i386/constraints.md (Yd): New constraint.
30075         (Ye): Ditto.
30076         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30077         and (?Yd, r) alternatives.  Update insn attributes.
30078         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30079         and (?*Yd, r) alternatives.  Update insn attributes.
30080         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30082 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30084         * gimplify.c (gimplify_modify_expr): Don't create a
30085         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30086         function.
30088 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30090         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30092 2017-05-30  Richard Biener  <rguenther@suse.de>
30094         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30095         and reduc_def fields.
30096         (STMT_VINFO_REDUC_TYPE): New define.
30097         (STMT_VINFO_REDUC_DEF): Likewise.
30098         (vect_force_simple_reduction): Adjust prototype.
30099         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30100         (vect_is_simple_reduction): Remove check_reduction argument.
30101         (vect_force_simple_reduction): Adjust and set
30102         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30103         (vectorizable_reduction): Do not re-do reduction analysis
30104         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30105         * tree-parloops.c (gather_scalar_reductions): Adjust.
30107 2017-05-30  Richard Biener  <rguenther@suse.de>
30109         PR middle-end/80901
30110         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30111         split_edge code.
30113 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30115         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30116         Introduce unknown_misalignment parameter and remove vf.
30117         (vect_peeling_hash_get_lowest_cost):
30118         Pass unknown_misalignment parameter.
30119         (vect_enhance_data_refs_alignment):
30120         Fix unsupportable data ref treatment.
30122 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30124         * tree-vect-data-refs.c (vect_get_data_access_cost):
30125         Workaround for SLP handling.
30126         (vect_enhance_data_refs_alignment):
30127         Compute costs for doing no peeling at all, compare to the best
30128         peeling costs so far and avoid peeling if cheaper.
30130 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30132         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30133         Return peeling info and set costs to zero for unlimited cost
30134         model.
30135         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30136         with unknown misalignment. Compute and costs for unknown
30137         misalignment, compare them to the costs for known misalignment
30138         and choose the cheapest for peeling.
30140 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30142         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30143         (vect_get_peeling_costs_all_drs): Create function.
30144         (vect_peeling_hash_get_lowest_cost):
30145         Use vect_get_peeling_costs_all_drs.
30146         (vect_peeling_supportable): Create function.
30147         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30149 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30151         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30152         DR_HAS_NEGATIVE_STEP.
30153         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30154         (vect_enhance_data_refs_alignment): Use.
30155         (vect_duplicate_ssa_name_ptr_info): Use.
30156         * tree-vectorizer.h (dr_misalignment): Use.
30157         (known_alignment_for_access_p): Use.
30159 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30161         PR target/78838
30162         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30163         .lowtext.
30164         (has_section_name): New function.
30166 2017-05-30  Martin Liska  <mliska@suse.cz>
30168         PR other/80909
30169         * auto-profile.c (get_function_decl_from_block): Fix
30170         parenthesis.
30172 2017-05-30  Richard Biener  <rguenther@suse.de>
30174         PR middle-end/80876
30175         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30177 2017-05-30  Martin Liska  <mliska@suse.cz>
30179         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30180         * dumpfile.h (struct dump_file_info): Remove ctors.
30182 2017-05-30  Martin Liska  <mliska@suse.cz>
30184         * predict.def: Fix GNU coding style.
30186 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30188         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30189         Mark 'to' argument with ATTRIBUTE_UNUSED.
30191 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30193         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30194         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30195         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30196         format string.
30198 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30200         * doc/install.texi (Options specification): Restore entry of
30201         --enable-sjlj-exceptions.
30203 2017-05-27  Michael Eager  <eager@eagercon.com>
30205         Revert:
30206         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30208         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30210         * config/microblaze/microblaze.h
30211         (FIXED_REGISTERS): Update in macro.
30212         (CALL_USED_REGISTERS): Update in macro.
30214 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30216         * doc/install.texi: Add links to macOS binary distributions.
30218 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30220         PR bootstrap/80887
30221         Revert:
30222         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30224         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30226 2017-05-26  Martin Liska  <mliska@suse.cz>
30228         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30230 2017-05-26  Martin Liska  <mliska@suse.cz>
30232         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30233         always leading ';; '.
30234         (dump_bb_info): Likewise.
30235         (brief_dump_cfg): Likewise.
30236         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30237         * dumpfile.c: Remove usage of TDF_VERBOSE.
30238         * dumpfile.h (enum dump_kind): Likewise.
30239         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30240         * print-tree.c (debug_verbose): Remove.
30241         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30242         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30243         * tree-diagnostic.c (default_tree_printer): Replace
30244         TDF_DIAGNOSTIC with TDF_SLIM.
30246 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30248         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30249         in parameter loop, rather than loop_vinfo.
30250         (create_intersect_range_checks): Ditto.
30251         (vect_create_cond_for_alias_checks): Update call to above functions.
30253 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30255         PR tree-optimization/80815
30256         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30257         for merging runtime alias checks.  Handle negative DR_STEPs.
30259 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30261         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30262         Move from ...
30263         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30264         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30265         out code pruning runtime alias checks.
30266         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30267         factored out from above.
30268         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30269         Move from ...
30270         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30271         ... to here.
30272         (prune_runtime_alias_test_list): New decalaration.
30274 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30276         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30277         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30278         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30279         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30280         (operator==, comp_dr_with_seg_len_pair): Ditto.
30281         (vect_prune_runtime_alias_test_list): Ditto.
30283 2017-05-26  Martin Liska  <mliska@suse.cz>
30285         PR ipa/80663
30286         * params.def: Bound partial-inlining-entry-probability param.
30288 2017-05-26  Marek Polacek  <polacek@redhat.com>
30290         PR sanitizer/80875
30291         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30292         can be negated.
30294 2017-05-26  Richard Biener  <rguenther@suse.de>
30296         PR tree-optimization/80842
30297         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30298         value.
30300 2017-05-26  Richard Biener  <rguenther@suse.de>
30302         PR tree-optimization/80844
30303         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30305 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30307         * doc/md.texi (Machine Constraints): Update x86 family
30308         machine constraints section to match 'config/i386/constraints.md'.
30310 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30312         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30314 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30316         * doc/invoke.texi (--enable-languages): Update documentation.
30318 2017-05-25  Martin Liska  <mliska@suse.cz>
30320         * dumpfile.c: Add TDF_FOLDING.
30321         * dumpfile.h (enum dump_kind): Likewise.
30322         * genmatch.c (dt_simplify::gen_1): Use it.
30324 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30326         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30328 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30330         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30331         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30333 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30335         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30336         * match.pd (X == C): Rewrite it here.
30337         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30338         with_certain_nonzero_bits2): New predicates.
30339         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30341 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30343         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30344         avoid warning.
30346         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30347         warning.
30349 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30351         * config/powerpcspe: New port.  Files are copied from the rs6000
30352         port, with "rs6000" in filenames replaced by "powerpcspe".
30354 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30356         PR rtl-optimization/80754
30357         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30359 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30361         * config/sparc/sparc.md (length): Return the correct value for -mflat
30362         sibcalls to match output_sibcall.
30364 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30366         PR bootstrap/80860
30367         PR bootstrap/80843
30368         * config/rs6000/rs6000.c (struct machine_function): Add new field
30369         n_components.
30370         (rs6000_get_separate_components): Init that field, use it.
30371         (rs6000_components_for_bb): Use the field.
30373 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30375         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30377 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30379         PR middle-end/80823
30380         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30382 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30384         PR target/80725
30385         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30386         address against address_operand predicate.
30387         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30389 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30391         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30392         parameters passed indirectly.
30394 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30396         * config/i386/i386.md (*movdi_internal): Remove SSE4
30397         alternative 18 (?r, *v).  Update insn attributes.
30398         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30399         Update insn attributes.
30400         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30401         Update insn attributes.
30402         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30403         alternative 1 (r, v). Remove isa attribute.
30404         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30405         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30406         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30408 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30410         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30411         dg-line directive.
30413 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30415         * cgraphunit.c (symbol_table::process_new_functions): Update.
30416         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30417         (inline_generate_summary): Rename to ...
30418         (ipa_fn_summary_generate): ... this one.
30419         (inline_read_summary): Rename to ...
30420         (ipa_fn_summary_read): ... this one.
30421         (inline_write_summary): Rename to ...
30422         (ipa_fn_summary_write): ... this one.
30423         (inline_free_summary): Rename to ...
30424         (ipa_free_fn_summary): ... this one.
30425         (pass_data_local_fn_summary, pass_local_fn_summary,
30426         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30427         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30428         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30429         make_pass_ipa_fn_summary): New.
30430         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30431         inline_write_summary, inline_free_summary): Remove.
30432         (ipa_free_fn_summary) : New.
30433         * ipa-inline.c (ipa_inline): Update.
30434         (pass_ipa_inline): Do not generate summaries.
30435         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30436         Remove.
30437         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30438         and add pass_ipa_fn_summary.
30439         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30440         New.
30441         (make_pass_inline_parameters): Remove.
30443 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30445         * omp-low.c (struct omp_context): Remove "default_kind" member.
30446         Adjust all users.
30448         * omp-offload.c (execute_oacc_device_lower): Remove the
30449         parallelism dimensions function attributes for unparallelized
30450         OpenACC kernels constructs.
30452 2017-05-23  Martin Liska  <mliska@suse.cz>
30454         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30455         functions.
30456         (cgraph_edge::make_speculative): Likewise.
30457         (cgraph_edge::resolve_speculation): Likewise.
30458         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30459         (cgraph_node::dump): Likewise.
30460         * cgraph.h: Likewise.
30461         * cgraphunit.c (analyze_functions): Likewise.
30462         (symbol_table::compile): Likewise.
30463         * ipa-cp.c (print_all_lattices): Likewise.
30464         (determine_versionability): Likewise.
30465         (initialize_node_lattices): Likewise.
30466         (ipcp_verify_propagated_values): Likewise.
30467         (estimate_local_effects): Likewise.
30468         (update_profiling_info): Likewise.
30469         (create_specialized_node): Likewise.
30470         (perhaps_add_new_callers): Likewise.
30471         (decide_about_value): Likewise.
30472         (decide_whether_version_node): Likewise.
30473         (identify_dead_nodes): Likewise.
30474         (ipcp_store_bits_results): Likewise.
30475         * ipa-devirt.c (dump_targets): Likewise.
30476         (ipa_devirt): Likewise.
30477         * ipa-icf.c (sem_item::dump): Likewise.
30478         (sem_function::equals): Likewise.
30479         (sem_variable::equals): Likewise.
30480         (sem_item_optimizer::read_section): Likewise.
30481         (sem_item_optimizer::execute): Likewise.
30482         (congruence_class::dump): Likewise.
30483         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30484         (dump_inline_summary): Likewise.
30485         (estimate_node_size_and_time): Likewise.
30486         (inline_analyze_function): Likewise.
30487         * ipa-inline-transform.c (inline_call): Likewise.
30488         * ipa-inline.c (report_inline_failed_reason): Likewise.
30489         (want_early_inline_function_p): Likewise.
30490         (edge_badness): Likewise.
30491         (update_edge_key): Likewise.
30492         (inline_small_functions): Likewise.
30493         * ipa-profile.c (ipa_profile): Likewise.
30494         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30495         (ipa_make_edge_direct_to_target): Likewise.
30496         (remove_described_reference): Likewise.
30497         (ipa_impossible_devirt_target): Likewise.
30498         (propagate_controlled_uses): Likewise.
30499         (ipa_print_node_params): Likewise.
30500         (ipcp_transform_function): Likewise.
30501         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30502         (propagate_pure_const): Likewise.
30503         * ipa-reference.c (generate_summary): Likewise.
30504         (read_write_all_from_decl): Likewise.
30505         (propagate): Likewise.
30506         (ipa_reference_read_optimization_summary): Likewise.
30507         * ipa-utils.c (ipa_merge_profiles): Likewise.
30508         * ipa.c (walk_polymorphic_call_targets): Likewise.
30509         (symbol_table::remove_unreachable_nodes): Likewise.
30510         (ipa_single_use): Likewise.
30511         * passes.c (execute_todo): Likewise.
30512         * predict.c (drop_profile): Likewise.
30513         * symtab.c (symtab_node::get_dump_name): New function.
30514         (symtab_node::dump_name): Likewise.
30515         (symtab_node::dump_asm_name): Likewise.
30516         (symtab_node::dump_references): Likewise.
30517         (symtab_node::dump_referring): Likewise.
30518         (symtab_node::dump_base): Likewise.
30519         (symtab_node::debug_symtab): Likewise.
30520         * tree-sra.c (convert_callers_for_node): Likewise.
30521         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30522         * value-prof.c (init_node_map): Likewise.
30524 2017-05-23  Martin Liska  <mliska@suse.cz>
30526         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30527         and symtab_node::debug_symtab to symbol_table::debug.
30528         * cgraphunit.c (analyze_functions): Use the renamed function.
30529         (symbol_table::compile): Likewise.
30530         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30531         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30532         * passes.c (execute_todo): Likewise.
30533         * symtab.c (symbol_table::dump): New function.
30534         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30536 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30538         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30539         that nonconst implies exec.
30541 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30543         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30544         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30545         (inline_edge_summary_vec): Turn into ...
30546         (ipa_call_summaries): ... this one.
30547         (redirect_to_unreachable, edge_set_predicate,
30548         evaluate_properties_for_edge, inline_summary_alloc,
30549         reset_ipa_call_summary, reset_inline_summary,
30550         inline_summary_t::duplicate): Update.
30551         (inline_edge_duplication_hook): Turn to ...
30552         (ipa_call_summary_t::duplicate): ... this one.
30553         (inline_edge_removal_hook): Turn to ...
30554         (ipa_call_summary_t::remove): ... this one.
30555         (dump_inline_edge_summary): Turn to ...
30556         (dump_ipa_call_summary): ... this one.
30557         (estimate_function_body_sizes): Update.
30558         (inline_update_callee_summaries): Update.
30559         (remap_edge_change_prob): Update.
30560         (remap_edge_summaries): Update.
30561         (inline_merge_summary): Update.
30562         (do_estimate_edge_time): Update.
30563         (inline_generate_summary): Update.
30564         (inline_read_section): Update.
30565         (inline_read_summary): Update.
30566         (inline_free_summary): Update.
30567         * ipa-inline.c (can_inline_edge_p): Update.
30568         (compute_inlined_call_time): Update.
30569         (want_inline_small_function_p): Update.
30570         (edge_badness): Update.
30571         (early_inliner): Update.
30572         * ipa-inline.h (inline_edge_summary): Turn to ...
30573         (ipa_call_summary): ... this one.
30574         (ipa_call_summary_t): New class.
30575         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30576         (ipa_call_summaries): New.
30577         (inline_edge_summary): Remove.
30578         (estimate_edge_growth): Update.
30579         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30580         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30581         * ipa-split.c (execute_split_functions): Update.
30582         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30584 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30586         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30587         attributes): Document rdrand effective target.
30589 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30591         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30592         attributes): Sort alphabetically.
30594 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30596         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30598 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30600         PR target/80718
30601         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30602         V2DF/V2DI splat into two separate patterns, one that handles
30603         registers, and the other that only handles memory.  Drop support
30604         for splatting from a GPR on ISA 2.07 and then splitting the
30605         splat into direct move and splat.
30606         (vsx_splat_<mode>_reg): Likewise.
30607         (vsx_splat_<mode>_mem): Likewise.
30609 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30611         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30613 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30615         PR middle-end/80809
30616         * omp-low.c (finish_taskreg_remap): New function.
30617         (finish_taskreg_scan): If unit size of ctx->record_type
30618         is non-constant, unshare the size expression and replace
30619         decls in it with possible outer var refs.
30621         PR middle-end/80809
30622         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30623         GOVD_SHARED rather than GOVD_PRIVATE with it.
30624         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30625         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30627         PR middle-end/80853
30628         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30629         as last argument to build_outer_var_ref for pointer bases of array
30630         section reductions.
30632 2017-05-19  Martin Sebor  <msebor@redhat.com>
30634         * print-tree.c (print_node): Print DECL_READ_P flag.
30636 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30638         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30639         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30640         * cgraph.c: Likewise.
30641         * cgraphunit.c: Likewise.
30642         * gengtype.c: Likewise.
30643         * ipa-cp.c: Likewise.
30644         * ipa-devirt.c: Likewise.
30645         * ipa-icf.c: Likewise.
30646         * ipa-predicate.c: Likewise.
30647         * ipa-profile.c: Likewise.
30648         * ipa-prop.c: Likewise.
30649         * ipa-split.c: Likewise.
30650         * ipa.c: Likewise.
30651         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30652         edge_predicate_pool, dump_inline_hints,
30653         inline_summary::account_size_time, redirect_to_unreachable,
30654         edge_set_predicate, set_hint_predicate,
30655         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30656         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30657         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30658         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30659         ipa_call_summary_t::remove, initialize_growth_caches,
30660         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30661         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30662         mark_modified, unmodified_parm_1, unmodified_parm,
30663         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30664         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30665         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30666         will_be_nonconstant_predicate, record_modified_bb_info,
30667         get_minimal_bb, record_modified, param_change_prob,
30668         phi_result_unknown_predicate, predicate_for_phi_result,
30669         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30670         estimate_function_body_sizes, compute_inline_parameters,
30671         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30672         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30673         inline_update_callee_summaries, remap_edge_change_prob,
30674         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30675         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30676         inline_analyze_function, inline_summary_t::insert,
30677         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30678         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30679         inline_free_summary): Move to ipa-fnsummary.h
30680         (predicate_t): Remove.
30681         * ipa-fnsummary.c: New file.
30682         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30683         (enum inline_hints_vals, inline_hints, agg_position_info,
30684         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30685         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30686         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30687         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30688         inline_read_summary, inline_write_summary, inline_free_summary,
30689         inline_analyze_function, initialize_inline_failed,
30690         inline_merge_summary, inline_update_overall_summary,
30691         compute_inline_parameters): Move to ipa-fnsummary.h
30692         * ipa-fnsummary.h: New file.
30693         * ipa-inline-transform.h: Include ipa-inline.h.
30694         * ipa-inline.c: LIkewise.
30696 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30698         * ipa-inline.c (edge_badness): Use inlined_time instead of
30699         inline_summaries->get.
30701 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30703         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30705 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30707         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30708         (fdump-lang): Document 'raw' option.
30709         * dumpfile.h (TDI_tu): Delete.
30710         * dumpfile.c (dump_files): Remove translation-unit.
30711         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30713 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30715         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30716         command option from $(AWK) call.
30717         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30718         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30719         [FORMAT]: Remove handling of variable.
30720         * config/avr/t-multilib: Regenerate.
30722 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30724         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30725         self_time.
30726         (dump_inline_summary): Do not print self_time.
30727         (estimate_function_body_sizes): Do not set self_time.
30728         (compute_inline_parameters): Likewise.
30729         (inline_read_section, inline_write_summary): Do not stream self_time.
30730         * ipa-inline.h (inline_summary): Drop self_time.
30732 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30734         * ipa-inline-analysis.c (account_size_time): Rename to ...
30735         (inline_summary::account_size_time): ... this one.
30736         (reset_ipa_call_summary): Turn to ...
30737         (ipa_call_summary::reset): ... this one.
30738         (reset_inline_summary): Turn to ...
30739         (inline_summary::reset): ... this one.
30740         (inline_summary_t::remove): Update.
30741         (inline_summary_t::duplicate): Update.
30742         (ipa_call_summary_t::remove): Update.
30743         (dump_inline_summary): Update.
30744         (estimate_function_body_sizes): Update.
30745         (compute_inline_parameters): Update.
30746         (estimate_node_size_and_time): Update.
30747         (inline_merge_summary): Update.
30748         (inline_update_overall_summary): Update.
30749         (inline_read_section): Update.
30750         (inline_write_summary): Update.
30751         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30752         add account_size_time and reset member functions.
30753         (ipa_call_summary): Add reset function.
30754         * ipa-predicate.h (predicate::operator &): Constify.
30756 2017-05-22  Richard Biener  <rguenther@suse.de>
30758         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30760 2017-05-19  Jason Merrill  <jason@redhat.com>
30762         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30764 2017-05-19  Marek Polacek  <polacek@redhat.com>
30766         PR sanitizer/80800
30767         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30768         TYPE_OVERFLOW_WRAPS checks.
30770 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30772         * tree-core.h (enum omp_clause_default_kind): Add
30773         "OMP_CLAUSE_DEFAULT_PRESENT".
30774         * tree-pretty-print.c (dump_omp_clause): Handle it.
30775         * gimplify.c (enum gimplify_omp_var_data): Add
30776         "GOVD_MAP_FORCE_PRESENT".
30777         (gimplify_adjust_omp_clauses_1): Map it to
30778         "GOMP_MAP_FORCE_PRESENT".
30779         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30781         * gimplify.c (oacc_default_clause): Clarify.
30783 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30785         LANG_HOOK_REGISTER_DUMPS
30786         * toplev.c (general_init): Call register dump lang hook.
30787         * doc/invoke.texi: Document -fdump-lang option family.
30788         * dumpfile.c (dump_files): Remove class dump here.
30789         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30790         * dumpfile.h (tree_dump_index): Remove TDI_class.
30791         * langhooks-def.h (lhd_register_dumps): Declare.
30792         (LANG_HOOKS_REGISTER_DUMPS): Define.
30793         (LANG_HOOKS_INITIALIZER): Add it.
30794         * langhooks.c (lhd_register_dumps): Define.
30795         * langhooks.h (struct lang_hooks): Add register_dumps.
30797 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30799         * context.h (context::set_passes): New.
30800         * context.c (context::context): Do not create pass manager.
30801         * toplev.c (general_init): Create pass manager here.
30803 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30805         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30806         use this splitter if two add or or instructions would also work for
30807         the constant we want to generate.
30809 2017-05-19  Richard Biener  <rguenther@suse.de>
30811         PR build/80821
30812         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30813         predicate evaluation.
30815 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30817         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30818         add ctor.
30819         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30820         unsigned.
30822 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30824         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30825         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30826         (inline_edge_summary_vec): Turn into ...
30827         (ipa_call_summaries): ... this one.
30828         (redirect_to_unreachable, edge_set_predicate,
30829         evaluate_properties_for_edge, inline_summary_alloc,
30830         reset_ipa_call_summary, reset_inline_summary,
30831         inline_summary_t::duplicate): Update.
30832         (inline_edge_duplication_hook): Turn to ...
30833         (ipa_call_summary_t::duplicate): ... this one.
30834         (inline_edge_removal_hook): Turn to ...
30835         (ipa_call_summary_t::remove): ... this one.
30836         (dump_inline_edge_summary): Turn to ...
30837         (dump_ipa_call_summary): ... this one.
30838         (estimate_function_body_sizes): Update.
30839         (inline_update_callee_summaries): Update.
30840         (remap_edge_change_prob): Update.
30841         (remap_edge_summaries): Update.
30842         (inline_merge_summary): Update.
30843         (do_estimate_edge_time): Update.
30844         (inline_generate_summary): Update.
30845         (inline_read_section): Update.
30846         (inline_read_summary): Update.
30847         (inline_free_summary): Update.
30848         * ipa-inline.c (can_inline_edge_p): Update.
30849         (compute_inlined_call_time): Update.
30850         (want_inline_small_function_p): Update.
30851         (edge_badness): Update.
30852         (early_inliner): Update.
30853         * ipa-inline.h (inline_edge_summary): Turn to ...
30854         (ipa_call_summary): ... this one.
30855         (ipa_call_summary_t): New class.
30856         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30857         (ipa_call_summaries): New.
30858         (inline_edge_summary): Remove.
30859         (estimate_edge_growth): Update.
30860         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30861         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30862         * ipa-split.c (execute_split_functions): Update.
30863         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30865 2017-05-19  Richard Biener  <rguenther@suse.de>
30867         PR middle-end/80764
30868         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
30870 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
30872         * config/rs6000/rs6000.c (struct machine_function): Add field
30873         fpr_is_wrapped_separately.
30874         (rs6000_get_separate_components): Use 64 components.  Handle the
30875         new FPR components.
30876         (rs6000_components_for_bb): Handle the FPR components.
30877         (rs6000_emit_prologue_components): Handle the FPR components.
30878         (rs6000_emit_epilogue_components): Handle the FPR components.
30879         (rs6000_set_handled_components): Handle the FPR components.
30880         (rs6000_emit_prologue): Don't output prologue code for those FPRs
30881         that are already separately shrink-wrapped.
30882         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
30883         that are already separately shrink-wrapped.
30885 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
30887         PR target/80510
30888         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
30889         New predicate.
30891         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
30892         (define_peephole2 for Altivec d-form load): Add peepholes to catch
30893         cases where the register allocator uses a move and an offsettable
30894         memory operation to/from a FPR register on ISA 2.06/2.07.
30895         (define_peephole2 for Altivec d-form store): Likewise.
30897 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
30899         PR target/80799
30900         * config/i386/mmx.md (*mov<mode>_internal): Enable
30901         alternatives 11, 12, 13 and 14 also for 32bit targets.
30902         Remove alternatives 15, 16, 17 and 18.
30903         * config/i386/sse.md (vec_concatv2di): Change
30904         alternative (!x, *y) to (x, ?!*Yn).
30906 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
30908         * dumpfile.h (enum dump_kind): Remove stray comma.
30910 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30912         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
30913         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
30914         predicate::num_conditions
30915         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
30916         (CHANGED): turn into predicate::changed.
30917         (agg_position_info): Move to ipa-predicate.h
30918         (add_condition, predicate::add_clause, predicate::operator &=,
30919         predicate::or_with, predicate::evaluate, predicate::probability,
30920         dump_condition, dump_clause, predicate::dump,
30921         predicate::remap_after_duplication, predicate::remap_after_inlining,
30922         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
30923         (evaluate_conditions_for_known_args): Update.
30924         (set_cond_stmt_execution_predicate): Update.
30925         * ipa-inline.h: Include ipa-predicate.h
30926         (condition, inline_param_summary, conditions, agg_position_info,
30927         predicate): Move to ipa-predicate.h
30928         * ipa-predicate.c: New file.
30929         * ipa-predicate.h: New file.
30931 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
30933         * final.c (leaf_function_p): Check we are not in a sequence.
30935 2017-05-18  Martin Liska  <mliska@suse.cz>
30937         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
30938         * dumpfile.c (dump_register): Use new enum dump_kind.
30939         (get_dump_file_name): Likewise.
30940         (dump_enable_all): Likewise.
30941         (dump_switch_p_1): Likewise.
30942         (enable_rtl_dump_file): Remove usage of TDF_RTL.
30943         * dumpfile.h (enum dump_kind): New enum type.
30944         (struct dump_file_info): Create constructor and
30945         format fields and comments.
30946         * passes.c (pass_manager::register_one_dump_file):
30947         Use num dump_kind.
30948         * statistics.c (statistics_early_init): Likewise.
30949         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
30950         TDF_TREE with TDF_SLIM.
30951         (gather_memory_references_ref): Likewise.
30953 2017-05-18  Martin Liska  <mliska@suse.cz>
30955         * vec.h (struct vnull): Use it.
30957 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30959         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
30960         (true_predicate, false_predicate, true_predicate_p,
30961         false_predicate_p): Remove.
30962         (single_cond_predicate, not_inlined_predicate): Turn to member function
30963         in ipa-inline.h
30964         (add_condition): Update.
30965         (add_clause): Turn to...
30966         (predicate::add_clause): ... this one; update; allow passing NULL
30967         as parameter.
30968         (and_predicates): Turn to ...
30969         (predicate::operator &=): ... this one.
30970         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
30971         (or_predicates): Turn to ...
30972         (predicate::or_with): ... this one.
30973         (evaluate_predicate): Turn to ...
30974         (predicate::evaluate): ... this one.
30975         (predicate_probability): Turn to ...
30976         (predicate::probability): ... this one.
30977         (dump_condition): Update.
30978         (dump_predicate): Turn to ...
30979         (predicate::dump): ... this one.
30980         (account_size_time): Update.
30981         (edge_set_predicate): Update.
30982         (set_hint_predicate): UPdate.
30983         (evaluate_conditions_for_known_args): Update.
30984         (evaluate_properties_for_edge): Update.
30985         (remap_predicate_after_duplication): Turn to...
30986         (predicate::remap_after_duplication): ... this one.
30987         (remap_hint_predicate_after_duplication): Update.
30988         (inline_summary_t::duplicate): UPdate.
30989         (dump_inline_edge_summary): Update.
30990         (dump_inline_summary): Update.
30991         (set_cond_stmt_execution_predicate): Update.
30992         (set_switch_stmt_execution_predicate): Update.
30993         (compute_bb_predicates): Update.
30994         (will_be_nonconstant_expr_predicate): Update.
30995         (will_be_nonconstant_predicate): Update.
30996         (phi_result_unknown_predicate): Update.
30997         (predicate_for_phi_result): Update.
30998         (array_index_predicate): Update.
30999         (estimate_function_body_sizes): Update.
31000         (estimate_node_size_and_time): Update.
31001         (estimate_ipcp_clone_size_and_time): Update.
31002         (remap_predicate): Rename to ...
31003         (predicate::remap_after_inlining): ... this one.
31004         (remap_hint_predicate): Update.
31005         (inline_merge_summary): Update.
31006         (inline_update_overall_summary): Update.
31007         (estimate_size_after_inlining): Update.
31008         (read_predicate): Rename to ...
31009         (predicate::stream_in): ... this one.
31010         (read_inline_edge_summary): Update.
31011         (write_predicate): Rename to ...
31012         (predicate::stream_out): ... this one.
31013         (write_inline_edge_summary): Update.
31014         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31015         (clause_t): Turn to uint32_t
31016         (predicate): Turn to class; implement constructor and operators
31017         ==, !=, &
31018         (size_time_entry): Update.
31019         (inline_summary): Update.
31020         (inline_edge_summary): Update.
31022 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31024         * fold-const.c (fold_binary_loc): Move transformation...
31025         * match.pd (C - X CMP X): ... here.
31027 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31029         * config/sparc/sparc.c (sparc_option_override): Set function
31030         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31031         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31032         latency to 1.
31033         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31034         latency to 2.
31035         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31037 2017-05-18  Marek Polacek  <polacek@redhat.com>
31039         PR sanitizer/80797
31040         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31041         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31042         gimple_assign_load_p.
31044 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31046         PR middle-end/80692
31047         * real.c (do_compare): Give decimal_do_compare preference over
31048         comparing just the signs.
31050 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31052         * doc/md.texi (Canonicalization of Instructions): Describe the
31053         canonical form of instructions that inherently set a condition
31054         code register.
31056 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31058         PR middle-end/80775
31059         * tree-cfg.c: Move deletion of unreachable case statements to after
31060         the merging of consecutive case labels.
31062 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31064         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31065         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31066         restoring of callee-saved registers.
31068 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31070         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31071         * config/visium/visium.c (single_set_and_flags): Likewise.
31072         * config/visium/visium.md (Substitutions): Likewise.
31074 2017-05-17  Martin Liska  <mliska@suse.cz>
31076         * cfg.c: Introduce dump_flags_t type and
31077         use it instead of int type.
31078         * cfg.h: Likewise.
31079         * cfghooks.c: Likewise.
31080         * cfghooks.h (struct cfg_hooks): Likewise.
31081         * cfgrtl.c: Likewise.
31082         * cfgrtl.h: Likewise.
31083         * cgraph.c (cgraph_node::get_body): Likewise.
31084         * coretypes.h: Likewise.
31085         * domwalk.c: Likewise.
31086         * domwalk.h: Likewise.
31087         * dumpfile.c (struct dump_option_value_info): Likewise.
31088         (dump_enable_all): Likewise.
31089         (dump_switch_p_1): Likewise.
31090         (opt_info_switch_p): Likewise.
31091         * dumpfile.h (enum tree_dump_index): Likewise.
31092         (struct dump_file_info): Likewise.
31093         * genemit.c: Likewise.
31094         * generic-match-head.c: Likewise.
31095         * gengtype.c (open_base_files): Likewise.
31096         * gimple-pretty-print.c: Likewise.
31097         * gimple-pretty-print.h: Likewise.
31098         * graph.c (print_graph_cfg): Likewise.
31099         * graphite-scop-detection.c (dot_all_sese): Likewise.
31100         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31101         * loop-unroll.c (report_unroll): Likewise.
31102         * passes.c (pass_manager::register_one_dump_file): Likewise.
31103         * print-tree.c: Likewise.
31104         * statistics.c: Likewise.
31105         * tree-cfg.c: Likewise.
31106         * tree-cfg.h: Likewise.
31107         * tree-dfa.c: Likewise.
31108         * tree-dfa.h: Likewise.
31109         * tree-dump.c (dump_function): Likewise.
31110         * tree-dump.h (struct dump_info): Likewise.
31111         * tree-pretty-print.c: Likewise.
31112         * tree-pretty-print.h: Likewise.
31113         * tree-ssa-live.c: Likewise.
31114         * tree-ssa-live.h: Likewise.
31115         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31116         * tree-vect-loop.c: Likewise.
31117         * tree-vect-slp.c: Likewise.
31119 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31120             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31122         PR tree-optimization/80457
31123         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31124         of all arguments to a statement as scalar_to_vec operations.
31125         (vectorizable_call): Adjust call to vect_model_simple_cost for
31126         new parameter.
31127         (vectorizable_conversion): Likewise.
31128         (vectorizable_assignment): Likewise.
31129         (vectorizable_shift): Likewise.
31130         (vectorizable_operation): Likewise.
31131         (vectorizable_comparison): Likewise.
31132         (vect_is_simple_cond): Record the def types for operands.
31133         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31134         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31135         for statement argument count.
31137 2017-05-16  Carl Love  <cel@us.ibm.com>
31139         * config/rs6000/rs6000-c: Add support for built-in functions
31140         vector unsigned long long vec_bperm (vector unsigned long long,
31141                                              vector unsigned char)
31142         vector signed long long vec_mule (vector signed int,
31143                                           vector signed int)
31144         vector unsigned long long vec_mule (vector unsigned int,
31145                                             vector unsigned int)
31146         vector signed long long vec_mulo (vector signed int,
31147                                           vector signed int)
31148         vector unsigned long long vec_mulo (vector unsigned int,
31149                                             vector unsigned int)
31150         vector signed char vec_sldw (vector signed char,
31151                                      vector signed char,
31152                                      const int)
31153         vector unsigned char vec_sldw (vector unsigned char,
31154                                        vector unsigned char,
31155                                        const int)
31156         vector signed short vec_sldw (vector signed short,
31157                                       vector signed short,
31158                                       const int)
31159         vector unsigned short vec_sldw (vector unsigned short,
31160                                         vector unsigned short,
31161                                         const int)
31162         vector signed int vec_sldw (vector signed int,
31163                                     vector signed int,
31164                                     const int)
31165         vector unsigned int vec_sldw (vector unsigned int,
31166                                       vector unsigned int,
31167                                       const int)
31168         vector signed long long vec_sldw (vector signed long long,
31169                                           vector signed long long,
31170                                           const int)
31171         vector unsigned long long vec_sldw (vector unsigned long long,
31172                                             vector unsigned long long,
31173                                             const int)
31174         * config/rs6000/rs6000-c: Add support for built-in functions
31175         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31176         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31177         * doc/extend.texi: Update the built-in documentation for the
31178         new built-in functions.
31180 2017-05-16  Marek Polacek  <polacek@redhat.com>
31182         PR sanitizer/80536
31183         PR sanitizer/80386
31184         * tree.c (save_expr): Don't fold the expression.
31186 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31188         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31189         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31190         and (?*y,m).  Update insn attributes.
31192 2017-05-16  Martin Liska  <mliska@suse.cz>
31194         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31195         flags argument of print_gimple_stmt, print_gimple_expr,
31196         print_generic_stmt and print_generic_expr.
31197         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31198         * coretypes.h: Likewise.
31199         * except.c (dump_eh_tree): Likewise.
31200         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31201         * gimple-pretty-print.h: Likewise.
31202         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31203         (backprop::push_to_worklist): Likewise.
31204         (backprop::pop_from_worklist): Likewise.
31205         (backprop::process_use): Likewise.
31206         (backprop::intersect_uses): Likewise.
31207         (note_replacement): Likewise.
31208         * gimple-ssa-store-merging.c
31209         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31210         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31211         (pass_store_merging::execute): Likewise.
31212         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31213         (ssa_base_cand_dump_callback): Likewise.
31214         (dump_incr_vec): Likewise.
31215         (replace_refs): Likewise.
31216         (replace_mult_candidate): Likewise.
31217         (create_add_on_incoming_edge): Likewise.
31218         (create_phi_basis): Likewise.
31219         (insert_initializers): Likewise.
31220         (all_phi_incrs_profitable): Likewise.
31221         (introduce_cast_before_cand): Likewise.
31222         (replace_one_candidate): Likewise.
31223         * gimplify.c (gimplify_expr): Likewise.
31224         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31225         (set_rename): Likewise.
31226         (rename_uses): Likewise.
31227         (copy_loop_phi_nodes): Likewise.
31228         (add_close_phis_to_merge_points): Likewise.
31229         (copy_loop_close_phi_args): Likewise.
31230         (copy_cond_phi_args): Likewise.
31231         (graphite_copy_stmts_from_block): Likewise.
31232         (translate_pending_phi_nodes): Likewise.
31233         * graphite-poly.c (print_pdr): Likewise.
31234         (dump_gbb_cases): Likewise.
31235         (dump_gbb_conditions): Likewise.
31236         (print_scop_params): Likewise.
31237         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31238         (build_cross_bb_scalars_use): Likewise.
31239         (gather_bbs::before_dom_children): Likewise.
31240         * hsa-dump.c (dump_hsa_immed): Likewise.
31241         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31242         (get_replacement_map): Likewise.
31243         * ipa-inline-analysis.c (dump_condition): Likewise.
31244         (estimate_function_body_sizes): Likewise.
31245         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31246         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31247         * ipa-prop.c (ipa_dump_param): Likewise.
31248         (ipa_print_node_jump_functions_for_edge): Likewise.
31249         (ipa_modify_call_arguments): Likewise.
31250         (ipa_modify_expr): Likewise.
31251         (ipa_dump_param_adjustments): Likewise.
31252         (ipa_dump_agg_replacement_values): Likewise.
31253         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31254         * ipa-pure-const.c (check_stmt): Likewise.
31255         (pass_nothrow::execute): Likewise.
31256         * ipa-split.c (execute_split_functions): Likewise.
31257         * omp-offload.c (dump_oacc_loop_part): Likewise.
31258         (dump_oacc_loop): Likewise.
31259         * trans-mem.c (tm_log_emit): Likewise.
31260         (tm_memopt_accumulate_memops): Likewise.
31261         (dump_tm_memopt_set): Likewise.
31262         (dump_tm_memopt_transform): Likewise.
31263         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31264         (print_loop): Likewise.
31265         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31266         (chkp_gather_checks_info): Likewise.
31267         (chkp_get_check_result): Likewise.
31268         (chkp_remove_check_if_pass): Likewise.
31269         (chkp_use_outer_bounds_if_possible): Likewise.
31270         (chkp_reduce_bounds_lifetime): Likewise.
31271         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31272         (chkp_mark_completed_bounds): Likewise.
31273         (chkp_register_incomplete_bounds): Likewise.
31274         (chkp_mark_invalid_bounds): Likewise.
31275         (chkp_maybe_copy_and_register_bounds): Likewise.
31276         (chkp_build_returned_bound): Likewise.
31277         (chkp_get_bound_for_parm): Likewise.
31278         (chkp_build_bndldx): Likewise.
31279         (chkp_get_bounds_by_definition): Likewise.
31280         (chkp_generate_extern_var_bounds): Likewise.
31281         (chkp_get_bounds_for_decl_addr): Likewise.
31282         * tree-chrec.c (chrec_apply): Likewise.
31283         * tree-data-ref.c (dump_data_reference): Likewise.
31284         (dump_subscript): Likewise.
31285         (dump_data_dependence_relation): Likewise.
31286         (analyze_overlapping_iterations): Likewise.
31287         * tree-inline.c (expand_call_inline): Likewise.
31288         (tree_function_versioning): Likewise.
31289         * tree-into-ssa.c (dump_defs_stack): Likewise.
31290         (dump_currdefs): Likewise.
31291         (dump_names_replaced_by): Likewise.
31292         (dump_update_ssa): Likewise.
31293         (update_ssa): Likewise.
31294         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31295         * tree-parloops.c (build_new_reduction): Likewise.
31296         (try_create_reduction_list): Likewise.
31297         (ref_conflicts_with_region): Likewise.
31298         (oacc_entry_exit_ok_1): Likewise.
31299         (oacc_entry_exit_single_gang): Likewise.
31300         * tree-pretty-print.h: Likewise.
31301         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31302         (get_scalar_evolution): Likewise.
31303         (add_to_evolution): Likewise.
31304         (get_loop_exit_condition): Likewise.
31305         (analyze_evolution_in_loop): Likewise.
31306         (analyze_initial_condition): Likewise.
31307         (analyze_scalar_evolution): Likewise.
31308         (instantiate_scev): Likewise.
31309         (number_of_latch_executions): Likewise.
31310         (gather_chrec_stats): Likewise.
31311         (final_value_replacement_loop): Likewise.
31312         (scev_const_prop): Likewise.
31313         * tree-sra.c (dump_access): Likewise.
31314         (disqualify_candidate): Likewise.
31315         (create_access): Likewise.
31316         (reject): Likewise.
31317         (maybe_add_sra_candidate): Likewise.
31318         (create_access_replacement): Likewise.
31319         (analyze_access_subtree): Likewise.
31320         (analyze_all_variable_accesses): Likewise.
31321         (sra_modify_assign): Likewise.
31322         (initialize_constant_pool_replacements): Likewise.
31323         (find_param_candidates): Likewise.
31324         (decide_one_param_reduction): Likewise.
31325         (replace_removed_params_ssa_names): Likewise.
31326         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31327         * tree-ssa-copy.c (dump_copy_of): Likewise.
31328         (copy_prop_visit_cond_stmt): Likewise.
31329         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31330         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31331         (record_equivalences_from_stmt): Likewise.
31332         * tree-ssa-dse.c (compute_trims): Likewise.
31333         (delete_dead_call): Likewise.
31334         (delete_dead_assignment): Likewise.
31335         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31336         (forward_propagate_into_cond): Likewise.
31337         (pass_forwprop::execute): Likewise.
31338         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31339         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31340         Likewise.
31341         (move_computations_worker): Likewise.
31342         (execute_sm): Likewise.
31343         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31344         (remove_exits_and_undefined_stmts): Likewise.
31345         (remove_redundant_iv_tests): Likewise.
31346         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31347         (adjust_iv_update_pos): Likewise.
31348         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31349         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31350         (value_replacement): Likewise.
31351         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31352         * tree-ssa-pre.c (print_pre_expr): Likewise.
31353         (get_representative_for): Likewise.
31354         (create_expression_by_pieces): Likewise.
31355         (insert_into_preds_of_block): Likewise.
31356         (eliminate_insert): Likewise.
31357         (eliminate_dom_walker::before_dom_children): Likewise.
31358         (eliminate): Likewise.
31359         (remove_dead_inserted_code): Likewise.
31360         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31361         * tree-ssa-reassoc.c (get_rank): Likewise.
31362         (eliminate_duplicate_pair): Likewise.
31363         (eliminate_plus_minus_pair): Likewise.
31364         (eliminate_not_pairs): Likewise.
31365         (undistribute_ops_list): Likewise.
31366         (eliminate_redundant_comparison): Likewise.
31367         (update_range_test): Likewise.
31368         (optimize_range_tests_var_bound): Likewise.
31369         (optimize_vec_cond_expr): Likewise.
31370         (rewrite_expr_tree): Likewise.
31371         (rewrite_expr_tree_parallel): Likewise.
31372         (linearize_expr): Likewise.
31373         (break_up_subtract): Likewise.
31374         (linearize_expr_tree): Likewise.
31375         (attempt_builtin_powi): Likewise.
31376         (attempt_builtin_copysign): Likewise.
31377         (transform_stmt_to_copy): Likewise.
31378         (transform_stmt_to_multiply): Likewise.
31379         (dump_ops_vector): Likewise.
31380         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31381         (print_scc): Likewise.
31382         (set_ssa_val_to): Likewise.
31383         (visit_reference_op_store): Likewise.
31384         (visit_use): Likewise.
31385         (sccvn_dom_walker::before_dom_children): Likewise.
31386         (run_scc_vn): Likewise.
31387         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31388         Likewise.
31389         (expr_hash_elt::print): Likewise.
31390         (const_and_copies::pop_to_marker): Likewise.
31391         (const_and_copies::record_const_or_copy_raw): Likewise.
31392         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31393         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31394         (dump_predicates): Likewise.
31395         (find_uninit_use): Likewise.
31396         (warn_uninitialized_phi): Likewise.
31397         (pass_late_warn_uninitialized::execute): Likewise.
31398         * tree-ssa.c (verify_vssa): Likewise.
31399         (verify_ssa): Likewise.
31400         (maybe_optimize_var): Likewise.
31401         * tree-vrp.c (dump_value_range): Likewise.
31402         (dump_all_value_ranges): Likewise.
31403         (dump_asserts_for): Likewise.
31404         (register_edge_assert_for_2): Likewise.
31405         (vrp_visit_cond_stmt): Likewise.
31406         (vrp_visit_switch_stmt): Likewise.
31407         (vrp_visit_stmt): Likewise.
31408         (vrp_visit_phi_node): Likewise.
31409         (simplify_cond_using_ranges_1): Likewise.
31410         (fold_predicate_in): Likewise.
31411         (evrp_dom_walker::before_dom_children): Likewise.
31412         (evrp_dom_walker::push_value_range): Likewise.
31413         (evrp_dom_walker::pop_value_range): Likewise.
31414         (execute_early_vrp): Likewise.
31416 2017-05-16  Richard Biener  <rguenther@suse.de>
31418         * dwarf2out.c (loc_list_from_tree_1): Do not create
31419         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31421 2017-05-16  Richard Biener  <rguenther@suse.de>
31423         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31424         just generated.
31425         (note_variable_value_in_expr): If we resolved the decl ref
31426         do not push to the stack.
31428 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31430         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31431         operations in fast-math mode.
31432         (vaddq_f16): Likewise.
31433         (vmul_f16): Likewise.
31434         (vmulq_f16): Likewise.
31435         (vsub_f16): Likewise.
31436         (vsubq_f16): Likewise.
31437         * config/arm/neon.md (add<mode>3): New.
31438         (sub<mode>3): New.
31439         (fma:<VH:mode>3): New.  Also remove outdated comment.
31440         (mul<mode>3): New.
31442 2017-05-16  Martin Liska  <mliska@suse.cz>
31444         PR ipa/79849.
31445         PR ipa/79850.
31446         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31447         (odr_types_equivalent_p): Likewise.
31449 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31451         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31453 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31455         PR target/80425
31456         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31457         non-interunit SSE move alternatives with '?'.
31458         (zero-extendsidi peephole2): New peephole to skip intermediate
31459         general register in SSE zero-extend sequence.
31461 2017-05-15  Jeff Law  <law@redhat.com>
31463         * reorg.c (relax_delay_slots): Create a new variable to hold
31464         the temporary target rather than clobbering TARGET_LABEL.
31466         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31467         missing argument to extract_bit_field call.
31468         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31470 2017-05-15  Martin Liska  <mliska@suse.cz>
31472         PR driver/31468
31473         * gcc.c (process_command): Do not allow empty argument of -o option.
31475 2017-05-15  Renlin Li  <renlin.li@arm.com>
31477         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31478         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31479         * config/aarch64/constraints.md (Usf): Add long call check.
31480         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31481         (call_value): Likewise.
31482         (sibcall): Likewise.
31483         (sibcall_value): Likewise.
31484         (call_insn): New.
31485         (call_value_insn): New.
31486         (sibcall_insn): Update rtx pattern.
31487         (sibcall_value_insn): Likewise.
31488         (call_internal): Remove.
31489         (call_value_internal): Likewise.
31490         (sibcall_internal): Likewise.
31491         (sibcall_value_internal): Likewise.
31492         (call_reg): Likewise.
31493         (call_symbol): Likewise.
31494         (call_value_reg): Likewise.
31495         (call_value_symbol): Likewise.
31497 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31499         PR target/80600
31500         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31502 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31504         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31505         compatible with CCGOCmode and with CCZmode.
31507 2017-05-14  Martin Sebor  <msebor@redhat.com>
31509         PR middle-end/77671
31510         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31511         (gimple_fold_builtin_snprintf): Same.
31512         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31513         (gimple_fold_builtin_snprintf): Same.
31514         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31515         of character types.
31516         (is_call_safe): New function.
31517         (try_substitute_return_value): Call it.
31518         (try_simplify_call): New function.
31519         (pass_sprintf_length::handle_gimple_call): Call it.
31521 2017-05-14  Martin Sebor  <msebor@redhat.com>
31523         PR middle-end/80669
31524         * builtins.c (expand_builtin_stpncpy): Simplify.
31526 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31528         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31529         * config/i386/i386.h
31530         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31531         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31532         (struct machine_function): Add new members call_ms2sysv,
31533         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31534         (struct machine_frame_state): New fields sp_realigned and
31535         sp_realigned_offset.
31536         * config/i386/i386.c
31537         (enum xlogue_stub): New enum.
31538         (enum xlogue_stub_sets): New enum.
31539         (class xlogue_layout): New class.
31540         (struct ix86_frame): New fields stack_realign_allocate_offset,
31541         stack_realign_offset and outlined_save_offset.  Modify comments to
31542         detail stack layout when using out-of-line stubs.
31543         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31544         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31545         -mcall-ms2sysv-xlogues.
31546         (stub_managed_regs): New static variable.
31547         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31548         registers managed by out-of-line stub.
31549         (disable_call_ms2sysv_xlogues): New function.
31550         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31551         m->call_ms2sysv when appropriate and compute frame layout for
31552         out-of-line stubs.
31553         (sp_valid_at, fp_valid_at): New inline functions.
31554         (choose_basereg): New function.
31555         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31556         all callers.
31557         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31558         Use align parameter of choose_baseaddr to generated aligned SSE movs
31559         when possible.
31560         (pro_epilogue_adjust_stack): Modify to track
31561         machine_frame_state::sp_realigned.
31562         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31563         (ix86_nsaved_sseregs): Likewise.
31564         (ix86_emit_save_regs): Likewise.
31565         (ix86_emit_save_regs_using_mov): Likewise.
31566         (ix86_emit_save_sse_regs_using_mov): Likewise.
31567         (get_scratch_register_on_entry): Likewise.
31568         (gen_frame_set): New function.
31569         (gen_frame_load): Likewise.
31570         (gen_frame_store): Likewise.
31571         (emit_outlined_ms2sysv_save): Likewise.
31572         (emit_outlined_ms2sysv_restore): Likewise.
31573         (ix86_expand_prologue): Modify stack re-alignment code and call
31574         emit_outlined_ms2sysv_save when appropriate.
31575         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31576         parameter rtx_insn *insn, which allows the function to be used to only
31577         generate the notes.
31578         (ix86_expand_epilogue): Modify validity checks of frame and stack
31579         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31580         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31581         * config/i386/predicates.md
31582         (save_multiple): New predicate.
31583         (restore_multiple): Likewise.
31584         * config/i386/sse.md
31585         (save_multiple<mode>): New pattern.
31586         (save_multiple_realign<mode>): Likewise.
31587         (restore_multiple<mode>): Likewise.
31588         (restore_multiple_and_return<mode>): Likewise.
31589         (restore_multiple_leave_return<mode>): Likewise.
31590         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31592 2017-05-14  Julia Koval  <julia.koval@intel.com>
31594         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31595         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31596         (__builtin_ia32_xsetbv): New builtins.
31597         * config/i386/i386.c (ix86_expand_special_args_builtin):
31598         Process new types.
31599         (ix86_expand_builtin): Special expand for new intrinsics.
31600         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31601         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31602         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31604 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31606         * cfganal.c (inverted_post_order_compute): Change argument type
31607         to vec *.
31608         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31609         * df-core.c (rest_of_handle_df_initialize): Adjust.
31610         (rest_of_handle_df_finish): Likewise.
31611         (df_analyze_1): Likewise.
31612         (df_analyze): Likewise.
31613         (loop_inverted_post_order_compute): Change argument to be a vec *.
31614         (df_analyze_loop): Adjust.
31615         (df_get_n_blocks): Likewise.
31616         (df_get_postorder): Likewise.
31617         * df.h (struct df_d): Change field to be a vec.
31618         * lcm.c (compute_laterin): Adjust.
31619         (compute_available): Likewise.
31620         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31621         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31622         * tree-ssa-pre.c (compute_antic): Likewise.
31624 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31626         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31627         (depth_first_search::depth_first_search): Change structure init
31628         function to this constructor.
31629         (depth_first_search::add_bb): Rename function to this member.
31630         (depth_first_search::execute): Likewise.
31631         (flow_dfs_compute_reverse_finish): Adjust.
31633 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31635         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31636         (longest_simple_path): Likewise.
31637         * shrink-wrap.c (spread_components): Likewise.
31638         (disqualify_problematic_components): Likewise.
31639         (emit_common_heads_for_components): Likewise.
31640         (emit_common_tails_for_components): Likewise.
31641         (insert_prologue_epilogue_for_components): Likewise.
31643 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31645         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31646         auto_sbitmap.
31648 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31650         * df-core.c (df_set_blocks): Start using auto_bitmap.
31651         (df_compact_blocks): Likewise.
31652         * df-problems.c (df_rd_confluence_n): Likewise.
31653         * df-scan.c (df_insn_rescan_all): Likewise.
31654         (df_process_deferred_rescans): Likewise.
31655         (df_update_entry_block_defs): Likewise.
31656         (df_update_exit_block_uses): Likewise.
31657         (df_entry_block_bitmap_verify): Likewise.
31658         (df_exit_block_bitmap_verify): Likewise.
31659         (df_scan_verify): Likewise.
31660         * lra-constraints.c (lra_constraints): Likewise.
31661         (undo_optional_reloads): Likewise.
31662         (lra_undo_inheritance): Likewise.
31663         * lra-remat.c (calculate_gen_cands): Likewise.
31664         (do_remat): Likewise.
31665         * lra-spills.c (assign_spill_hard_regs): Likewise.
31666         (spill_pseudos): Likewise.
31667         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31668         (bitmap_set_subtract_values): Likewise.
31670 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31672         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31673         management with auto_bitmap.
31674         (fix_inter_tick): Likewise.
31675         (fix_recovery_deps): Likewise.
31676         * ira.c (add_store_equivs): Likewise.
31677         (find_moveable_pseudos): Likewise.
31678         (split_live_ranges_for_shrink_wrap): Likewise.
31679         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31680         (rtx_reuse_manager::seen_def_p): Likewise.
31681         (rtx_reuse_manager::set_seen_def): Likewise.
31682         * print-rtl.h (class rtx_reuse_manager): Likewise.
31684 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31686         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31687         lifetime.
31688         (migrate_btr_def): Likewise.
31689         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31690         * df-core.c (loop_post_order_compute): Likewise.
31691         (loop_inverted_post_order_compute): Likewise.
31692         * hsa-common.h: Likewise.
31693         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31694         * init-regs.c (initialize_uninitialized_regs): Likewise.
31695         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31696         (inline_small_functions): Likewise.
31697         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31698         * ira.c (combine_and_move_insns): Likewise.
31699         (build_insn_chain): Likewise.
31700         * loop-invariant.c (find_invariants): Likewise.
31701         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31702         * predict.c (tree_predict_by_opcode): Likewise.
31703         (predict_paths_leading_to): Likewise.
31704         (predict_paths_leading_to_edge): Likewise.
31705         (estimate_loops_at_level): Likewise.
31706         (estimate_loops): Likewise.
31707         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31708         (spread_components): Likewise.
31709         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31710         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31711         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31712         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31713         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31714         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31715         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31716         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31717         (mark_threaded_blocks): Likewise.
31718         (thread_through_all_blocks): Likewise.
31719         * tree-ssa.c (verify_ssa): Likewise.
31720         (execute_update_addresses_taken): Likewise.
31721         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31723 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31725         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31726         auto_vec.
31727         (post_order_compute): Likewise.
31728         (inverted_post_order_compute): Likewise.
31729         (pre_and_rev_post_order_compute_fn): Likewise.
31731 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31733         * genrecog.c (int_set::int_set): Explicitly construct our
31734         auto_vec base class.
31735         * vec.h (auto_vec::auto_vec): New constructor.
31737 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31739         * bitmap.h (class auto_bitmap): New constructor taking
31740         bitmap_obstack * argument.
31742 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31744         * bitmap.h (class auto_bitmap): Change type of m_bits to
31745         bitmap_head, and adjust ctor / dtor and member operators.
31747 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31749         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31750         when returned register mode doesn't match original mode.
31752 2017-05-12  Jeff Law  <law@redhat.com>
31753             Jakub Jelinek  <jakub@redhat.com>
31755         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31756         we look for cc setter after the compare-elim changes.
31757         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31758         within the vector to match what compare-elim now expects.
31759         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31760         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31762         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31763         after the compare-elim changes.
31764         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31765         the vector to match what compare-elim now expects.
31766         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31767         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31768         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31769         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31770         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31772         * config/visium/visium.c (single_set_and_flags): Fix where
31773         we look for cc setter after the compare-elim changes.
31774         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31775         with the vector to match what compare-elim now expects.
31776         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31777         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31778         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31779         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31780         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31781         (neg<mode>2_insn_set_overflow): Likewise.
31783 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31785         PR middle-end/79794
31786         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31787         maybe_expand_insn call, set ops[0].target.  If still set after call,
31788         set alt_rtl.  Add extra arg to recursive calls.
31789         (extract_bit_field): Add alt_rtl argument.  Pass to
31790         extract_bit_field.
31791         * expmed.h (extract_bit_field): Fix prototype.
31792         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31793         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31794         to extract_bit_field_calls.
31795         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31796         Pass alt_rtl to extract_bit_field calls.
31797         * calls.c (store_unaligned_arguments_into_psuedos)
31798         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31799         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31800         gen_reg_rtx.
31801         * optabs.h (struct expand_operand): Add target bitfield.
31803 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31805         * compare-elim.c (try_eliminate_compare): Canonicalize
31806         operation with embedded compare to
31807         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31808          (set (reg) (operation)].
31810         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31812 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31814         PR target/80723
31815         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31816         cost of adding a carry flag for ADC instruction.
31817         [case MINUS]: Ignore the cost of subtracting a carry flag
31818         for SBB instruction.
31820 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31822         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31823         and x86intrin.h
31824         * config/rs6000/bmiintrin.h: New file.
31825         * config/rs6000/bmi2intrin.h: New file.
31826         * config/rs6000/x86intrin.h: New file.
31828 2017-05-12  Jeff Law  <law@redhat.com>
31830         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31831         markers.
31833 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31835         PR middle-end/80707
31836         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31838 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31840         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31841         early expansion of vector divide builtins.
31842         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31843         builtins identified as having unsigned arguments.
31845 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31847         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31848         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31849         expansion of vector logical operations (and, andc, or, xor,
31850         nor, orc, nand).
31852 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31854         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31855         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31857 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31859         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31860         early GIMPLE expansion of vector multiplies.
31862 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
31864         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
31865         TARGET_HAVE_MOVT conditional.
31866         (movt splitter): Likewise.
31868 2017-05-12  Richard Biener  <rguenther@suse.de>
31870         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
31871         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
31872         Fold all stmts not inplace.
31874 2017-05-12  Richard Biener  <rguenther@suse.de>
31876         PR tree-optimization/80713
31877         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
31878         inserted_exprs bit for not removed stmts.
31880 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
31882         PR middle-end/69921
31883         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
31884         parallelized" attribute for parallelized OpenACC kernels.
31885         * omp-offload.c (execute_oacc_device_lower): Use it.
31887         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
31888         Set "oacc kernels" attribute.
31889         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
31890         parameter.  Adjust all users.
31891         (oacc_fn_attrib_kernels_p): Remove function.
31892         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
31893         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
31894         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
31895         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
31896         assert "oacc kernels" attribute is set.
31898 2017-05-11  Carl Love  <cel@us.ibm.com>
31900         * config/rs6000/rs6000-c: Add support for built-in functions
31901         vector unsigned char vec_popcnt (vector signed char)
31902         vector unsigned char vec_popcnt (vector unsigned char)
31903         vector unsigned short vec_popcnt (vector signed short)
31904         vector unsigned short vec_popcnt (vector unsigned short)
31905         vector unsigned int vec_popcnt (vector signed int)
31906         vector unsigned int vec_popcnt (vector unsigned int)
31907         vector unsigned long long vec_popcnt (vector signed long long)
31908         vector unsigned long long vec_popcnt (vector unsigned long long)
31909         vector signed long long vec_slo (vector signed long long,
31910                                          vector signed char)
31911         vector signed long long vec_slo (vector signed long long,
31912                                          vector unsigned char)
31913         vector unsigned long long vec_slo (vector unsigned long long,
31914                                            vector signed char)
31915         vector unsigned long long vec_slo (vector unsigned long long,
31916                                            vector unsigned char)
31917         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
31918         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
31919         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
31920         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
31921         * doc/extend.texi: Update the built-in documentation file for the
31922         new built-in functions.
31924 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
31926         * attribs.h (sorted_attr_string): Move machine independent
31927         functions for target clone support from the i386 port to common
31928         code.  Rename ix86_function_versions to common_function_versions.
31929         Rename make_name to make_unique_name.
31930         (common_function_versions): Likewise.
31931         (make_unique_name): Likewise.
31932         (make_dispatcher_decl): Likewise.
31933         (is_function_default_version): Likewise.
31934         * attribs.c (attr_strcmp): Likewise.
31935         (sorted_attr_string): Likewise.
31936         (common_function_versions): Likewise.
31937         (make_unique_name): Likewise.
31938         (make_dispatcher_decl): Likewise.
31939         (is_function_default_version): Likewise.
31940         * config/i386/i386.c (attr_strcmp): Likewise.
31941         (sorted_attr_string): Likewise.
31942         (ix86_function_versions): Likewise.
31943         (make_name): Likewise.
31944         (make_dispatcher_decl): Likewise.
31945         (is_function_default_version): Likewise.
31946         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
31948 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31950         PR target/80695
31951         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
31952         Account for direct move costs for vec_construct of integer
31953         vectors.
31955 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
31957         PR target/80706
31958         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
31959         (UNSPEC_STX_ATOMIC): Ditto.
31960         (loaddi_via_sse): New insn.
31961         (storedi_via_sse): Ditto.
31962         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
31963         Update corresponding peephole2 patterns.
31964         (atomic_storedi_fpu): Ditto.
31966 2017-05-11  Julia Koval  <julia.koval@intel.com>
31968         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
31969         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
31970         New intrinsics.
31971         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
31972         (__builtin_ia32_rsqrt14ss_mask): New builtins.
31973         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
31975 2017-05-11  Nathan Sidwell  <nathan@acm.org>
31977         * graphite-poly.c: Include dumpfile.h.
31979         * dumpfle.h (dump_function): Declare here ...
31980         * tree-dump.h (dump_function): ... not here.
31981         * dumpfile.c: #include tree-cfg.h.
31982         (dump_function): Move here from ...
31983         * tree-dump.c (dump_function): ... here.
31984         * gimplify.c: #include splay-tree.h, not tree-dump.h.
31985         * graphite-poly.c: Don't include tree-dump.h.
31986         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
31987         * print-tree.c: Likewise.
31988         * stor-layout.c: Likewise.
31989         * tree-nested.c: Likewise.
31991         * dumpfile.c (dump_start): Use TDF_FLAGS.
31992         (dump_enable_all): Fix TDF_KIND check thinko.
31994 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
31996         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
31997         array entries to represent two legal parameterizations of the
31998         overloaded __builtin_cmpb function, as represented by the
31999         P6_OV_BUILTIN_CMPB constant.
32000         (altivec_resolve_overloaded_builtin): Add special case handling
32001         for the __builtin_cmpb function, as represented by the
32002         P6_OV_BUILTIN_CMPB constant.
32003         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32004         (BU_P6_64BIT_2): New macro.
32005         (BU_P6_OVERLOAD_2): New macro
32006         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32007         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32008         (CMPB): Add overload support to represent both 32-bit and 64-bit
32009         compare-bytes function.
32010         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32011         support for TARGET_CMPB.
32012         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32013         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32014         documentation of the __builtin_cmpb overloaded built-in function.
32016 2017-05-11  Richard Biener  <rguenther@suse.de>
32018         PR tree-optimization/80705
32019         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32020         bases are not vectorizable.
32022 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32024         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32025         when counting register pressure.
32027 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32029         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32030         (iv_ca_more_deps): Renamed to ...
32031         (iv_ca_compare_deps): ... this.
32032         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32034 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32036         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32037         to ...
32038         (determine_group_iv_costs): ... here.
32039         (find_inv_vars_cb): Record inv var if it's not recorded before.
32041 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32043         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32044         (get_shiftadd_cost): Ditto.
32046 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32048         * tree-ssa-address.c: Include header file.
32049         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32050         address.
32051         (add_to_parts): Refactor.
32052         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32053         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32054         in new order.
32056 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32058         PR tree-optimization/53090
32059         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32060         COMP_IV_EXPR_2.
32061         (extract_cond_operands): Detect condition with IV on both sides
32062         and return COMP_IV_EXPR_2.
32063         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32064         (rewrite_use_compare): Simplify by removing call to function
32065         extract_cond_operands.
32067 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32069         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32070         (extract_cond_operands): Detect condition comparing against non-
32071         invariant bound and return appropriate enum value.
32072         (find_interesting_uses_cond): Update use of extract_cond_operands.
32073         Handle its return value accordingly.
32074         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32076 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32078         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32079         nonlinear iv_use computation in loop invariant sensitive way.
32081 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32083         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32084         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32086 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32088         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32089         (dump_cand): Support iv_cand.inv_exprs.
32090         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32091         for candidates.
32092         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32093         iv_cand.inv_exprs.
32095 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32097         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32098         from ...
32099         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32100         as local function.  Include necessary header files.
32101         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32103 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32105         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32107 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32109         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32110         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32111         RSHIFT_EXPR and BIT_NOT_EXPR.
32113 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32115         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32116         (adjust_setup_cost): New parameter supporting round up adjustment.
32117         (struct address_cost_data): Delete.
32118         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32119         (split_address_cost, ptr_difference_cost): Delete.
32120         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32121         (struct ainc_cost_data): New struct.
32122         (get_address_cost_ainc): New function.
32123         (get_address_cost, get_computation_cost): Reimplement.
32124         (determine_group_iv_cost_address): Record inv_expr for all uses of
32125         a group.
32126         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32127         (iv_ca_has_deps): Reimplemented to ...
32128         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32129         than OLD_CP.
32130         (iv_ca_extend): Call iv_ca_more_deps.
32132 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32134         * tree-ssa-address.c (struct mem_address): Move to header file.
32135         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32136         * tree-ssa-address.h (struct mem_address): Move from C file.
32137         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32139 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32141         * tree-affine.h (aff_combination_type): New interface.
32142         (aff_combination_zero_p): Remove static.
32143         (aff_combination_const_p): New interface.
32144         (aff_combination_singleton_var_p): New interfaces.
32146 2017-05-11  Richard Biener  <rguenther@suse.de>
32148         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32149         Skip unreachable blocks and destinations.
32150         (eliminate): Move stmt removal and fixup ...
32151         (fini_eliminate): ... here.  Skip inserted exprs.
32152         (pass_pre::execute): Move fini_pre after fini_eliminate.
32153         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32154         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32155         PRE to get rid of dead code that has invalid SSA form and
32156         split critical edges again.
32158 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32160         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32162 2017-05-11  Richard Biener  <rguenther@suse.de>
32164         * passes.c (execute_function_todo): Verify loops if they are
32165         said to be up-to-date.
32166         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32167         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32169 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32171         PR target/80090
32172         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32173         handle calling assemble_external ourself.
32175         PR target/79027
32176         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32177         modes with zero size.  Enhance comment.
32179 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32181         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32182         built-ins for vec_xl and vec_xst with short and char pointer
32183         arguments.
32185 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32187         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32188         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32189         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32190         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32191         (_mm_maskz_min_round_ss): New intrinsics.
32192         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32193         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32194         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32195         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32196         (__builtin_ia32_minss_mask_round): New builtins.
32197         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32198         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32199         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32200         Rename to ...
32201         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32202         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32203         Change to ...
32204         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32205         ... this.
32207 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32209         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32210         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32211         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32212         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32213         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32214         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32215         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32216         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32217         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32218         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32219         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32220         (__builtin_ia32_mulss_mask_round): New builtins.
32221         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32222         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32223         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32224         Rename to ...
32225         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32226         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32227         Change to ...
32228         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32229         ... this.
32231 2017-05-10  Julia Koval  <julia.koval@intel.com>
32233         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32234         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32235         (_mm256_setr_m128i): New intrinsics.
32237 2017-05-10  Julia Koval  <julia.koval@intel.com>
32239         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32240         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32241         (_mm_maskz_rcp14_ss): New intrinsics.
32242         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32243         (__builtin_ia32_rcp14ss_mask): New builtins.
32244         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32246 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32248         PR tree-optimization/51513
32249         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32250         (assert_unreachable_fallthru_edge_p): Use it.
32251         (group_case_labels_stmt): Likewise.
32252         * tree-cfg.h: Prototype it.
32253         * stmt.c: Include cfghooks.h and tree-cfg.h.
32254         (emit_case_dispatch_table) <gap_label>: New local variable.
32255         Use it to fill dispatch table gaps.
32256         Test for default_label before updating probabilities.
32257         (expand_case) <default_label>: Remove unneeded initialization.
32258         Test for unreachable default case statement and remove its edge.
32259         Set default_label accordingly.
32260         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32262 2017-05-10  Carl Love  <cel@us.ibm.com>
32264         * config/rs6000/rs6000-c: Add support for built-in functions
32265         vector signed char      vec_neg (vector signed char)
32266         vector signed short int vec_neg (vector short int)
32267         vector signed int       vec_neg (vector signed int)
32268         vector signed long long vec_neg (vector signed long long)
32269         vector float            vec_neg (vector float)
32270         vector double           vec_neg (vector double)
32271         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32272         overload.
32273         * config/rs6000/altivec.h: Add define for vec_neg
32274         * doc/extend.texi: Update the built-in documentation for the
32275         new built-in functions.
32277 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32279         PR tree-optimization/77644
32280         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32282 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32284         * dumpfile.h (TDI_lang_all): New.
32285         (TDF_KIND): New. Renumber others
32286         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32287         than bits.
32288         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32289         lang-all.
32290         (get_dump_file_name): Adjust suffix generation.
32291         (dump_enable_all): Use TDF_KIND.
32292         * doc/invoke.texi (-fdump-lang-all): Document.
32294         * dumpfile.h: Tabify.
32296 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32298         PR target/80671
32299         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32300         Move member access before delete.
32302 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32304         * tree-inline.c (expand_call_inline): Split block at stmt
32305         before the call.
32307 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32309         PR target/68163
32310         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32311         are now unused after splitting mov{sf,sd}_hardfloat.
32312         (f32_lr2): Likewise.
32313         (f32_lm): Likewise.
32314         (f32_lm2): Likewise.
32315         (f32_li): Likewise.
32316         (f32_li2): Likewise.
32317         (f32_lv): Likewise.
32318         (f32_sr): Likewise.
32319         (f32_sr2): Likewise.
32320         (f32_sm): Likewise.
32321         (f32_sm2): Likewise.
32322         (f32_si): Likewise.
32323         (f32_si2): Likewise.
32324         (f32_sv): Likewise.
32325         (f32_dm): Likewise.
32326         (f32_vsx): Likewise.
32327         (f32_av): Likewise.
32328         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32329         For movsf, order stores so the VSX stores occur before the GPR
32330         store which encourages the register allocator to use a traditional
32331         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32332         store comes before the VSX stores to allow the power6 to work.
32333         This is due to the power6 not having a 32-bit integer store
32334         instruction from a FPR.
32335         (movsf_hardfloat): Likewise.
32336         (movsd_hardfloat): Likewise.
32338 2017-05-09  Martin Sebor  <msebor@redhat.com>
32340         PR translation/80280
32341         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32342         added in r247778.
32344         PR translation/80280
32345         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32346         data member added in r247778.
32347         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32349 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32351         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32353         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32354         typedefs.
32356 2017-05-09  Marek Polacek  <polacek@redhat.com>
32358         * doc/invoke.texi: Fix typo.
32360 2017-05-09  Richard Biener  <rguenther@suse.de>
32362         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32363         (vrp_val_is_min): Likewise.
32364         (set_value_range_to_value): Likewise.
32365         (set_value_range_to_nonnegative): Likewise.
32366         (gimple_assign_nonzero_p): Likewise.
32367         (gimple_stmt_nonzero_p): Likewise.
32368         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32369         (adjust_range_with_scev): Adjust comments.
32370         (compare_range_with_value): Likewise.
32371         (extract_range_from_phi_node): Likewise.
32372         (test_for_singularity): Likewise.
32374 2017-05-09  Richard Biener  <rguenther@suse.de>
32376         * tree-vrp.c (get_single_symbol): Add assert that we don't
32377         get overflowed constants as invariant part.
32378         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32379         checks.  Use wi::cmp instead of recursing for integer constants.
32380         (compare_values): Just ignore whether we assumed undefined
32381         overflow instead of failing the compare.
32382         (extract_range_for_var_from_comparison_expr): Add comment before the
32383         TREE_NO_WARNING sets.
32384         (test_for_singularity): Likewise.
32385         (extract_range_from_comparison): Do not disable optimization
32386         when we assumed undefined overflow.
32387         (extract_range_basic): Remove init of unused var.
32389 2017-05-09  Richard Biener  <rguenther@suse.de>
32391         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32392         (extract_range_from_multiplicative_op_1): Adjust.
32393         (extract_range_from_binary_expr_1): Use int_const_binop.
32395 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32397         PR target/80101
32398         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32399         rs6000_store_data_bypass_p in seven define_bypass directives and
32400         in several comments.
32401         * config/rs6000/rs6000-protos.h: Add prototype for
32402         rs6000_store_data_bypass_p function.
32403         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32404         function implements slightly different (rs6000-specific) semantics
32405         than store_data_bypass_p, returning false rather than aborting
32406         with assertion error when arguments do not satisfy the
32407         requirements of store data bypass.
32408         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32409         rs6000_store_data_bypass_p.
32411 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32413         * config/xtensa/xtensa-protos.h
32414         (xtensa_initial_elimination_offset): New declaration.
32415         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32416         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32417         macro definition, add case for FRAME_POINTER_REGNUM when
32418         FRAME_GROWS_DOWNWARD.
32419         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32420         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32421         xtensa_initial_elimination_offset.
32423 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32425         * doc/invoke.texi: Alphabetize -fdump options.
32427 2017-05-08  Martin Sebor  <msebor@redhat.com>
32429         PR translation/80280
32430         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32432 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32434         * target.def (compute_frame_layout): New optional target hook.
32435         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32436         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32437         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32438         target hook.
32439         * reload1.c (verify_initial_elim_offsets): Likewise.
32440         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32441         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32442         (arm_get_frame_offsets): Split up into this ...
32443         (arm_compute_frame_layout): ... and this function.
32445 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32447         * config/aarch64/constraints.md (Usa): New constraint.
32448         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32450 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32452         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32453         with_multilib_list after it has been checked.
32455 2017-05-08  Richard Biener  <rguenther@suse.de>
32457         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32458         (bitmap_set_subtract_values): Likewise.
32460 2017-05-08  Richard Biener  <rguenther@suse.de>
32462         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32463         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32464         argument.
32465         (gimple_stmt_nonzero_warnv_p): Rename to ...
32466         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32467         argument.
32468         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32469         (extract_range_basic): Adjust, do not disable propagation on
32470         strict overflow sensitive simplification.
32471         (vrp_visit_cond_stmt): Likewise.
32473 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32475         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32476         body size unconditionally.
32478 2017-05-07  Jeff Law  <law@redhat.com>
32480         Revert:
32481         2017-05-06  Jeff Law  <law@redhat.com>
32482         PR tree-optimization/78496
32483         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32484         code.
32486         PR tree-optimization/78496
32487         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32488         (simplify_stmt_using_ranges): Call it.
32489         (vrp_dom_walker::before_dom_children): Extract equivalences
32490         from an ASSERT_EXPR with an equality comparison against a
32491         constant.
32493 2017-05-06  Jeff Law  <law@redhat.com>
32495         PR tree-optimization/78496
32496         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32497         code.
32499         PR tree-optimization/78496
32500         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32501         (simplify_stmt_using_ranges): Call it.
32502         (vrp_dom_walker::before_dom_children): Extract equivalences
32503         from an ASSERT_EXPR with an equality comparison against a
32504         constant.
32506 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32508         * lra-constraints.c (lra_copy_reg_equiv): New function.
32509         (split_reg): Use it to copy equivalence information from the
32510         original register to the spill register.
32512 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32514         PR rtl-optimization/75964
32515         * simplify-rtx.c (simplify_const_relational_operation): Remove
32516         invalid handling of comparisons of integer ABS.
32518 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32520         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32521         initialize to zero.
32522         (init_regs): Remove declaration.
32523         (function_arg_advance_32): Initialize error_p as boolean variable.
32525 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32527         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32528         lines.  Use for (;;).
32530 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32532         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32533         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32534         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32535         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32536         VF=2 that require versioning.
32538 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32540         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32541         int.
32543 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32545         * diagnostic.h (diagnostic_override_option_index): Convert from
32546         macro to inline function.
32548 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32550         * diagnostic.c (last_module_changed_p): New function.
32551         (set_last_module): New function.
32552         (diagnostic_report_current_module): Convert macro usage to
32553         the above functions.
32554         * diagnostic.h (diagnostic_context::last_module): Strengthen
32555         from const line_map * to const line_map_ordinary *.
32556         (diagnostic_last_module_changed): Delete macro.
32557         (diagnostic_set_last_module): Delete macro.
32559 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32561         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32562         with diagnostic_report_diagnostic.
32563         (diagnostic_n_impl_richloc): Likewise.
32564         * diagnostic.h (report_diagnostic): Delete macro.
32565         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32566         with diagnostic_report_diagnostic.
32567         * substring-locations.c (format_warning_va): Likewise.
32569 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32571         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32572         save/restor of format_spec.  Move option-printing code to...
32573         (print_option_information): ...this new function, and
32574         reimplement by simply printing to the pretty_printer,
32575         rather than appending to the format string.
32577 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32579         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32580         handling logic into...
32581         (update_effective_level_from_pragmas): ...this new function.
32583 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32585         * config/riscv/riscv.opt (mstrict-align): New option.
32586         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32587         (SLOW_UNALIGNED_ACCESS): Define.
32588         (riscv_slow_unaligned_access): Declare.
32589         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32590         field.
32591         (riscv_slow_unaligned_access): New variable.
32592         (rocket_tune_info): Set slow_unaligned_access to true.
32593         (optimize_size_tune_info): Set slow_unaligned_access to false.
32594         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32595         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32596         (riscv_option_override): Set riscv_slow_unaligned_access.
32597         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32599 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32601         * config/riscv/riscv.md: Unify indentation.
32603 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32605         PR target/79038
32606         PR target/79202
32607         PR target/79203
32608         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32609         UNSIGNED_FIX.
32610         (extendsi<mode>2): Add support for doing sign extension via
32611         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32612         don't have ISA 3.0 instructions.
32613         (extendsi<mode>2 splitter): Likewise.
32614         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32615         generate the normal insns since SImode can now go in vector
32616         registers.  Disallow the special UNSPECs needed for previous
32617         machines to hide SImode being used.  Add new insns
32618         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32619         (fix_trunc<mode>si2_stfiwx): Likewise.
32620         (fix_trunc<mode>si2_internal): Likewise.
32621         (fixuns_trunc<mode>si2): Likewise.
32622         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32623         (fctiw<u>z_<mode>_smallint): Likewise.
32624         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32625         of floating point to 32-bit integer from doing a direct move to
32626         the GPR registers to do a store.
32627         (fctiwz_<mode>): Break long line.
32629 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32631         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32632         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32633         (addr_list, addr_offset_valid_p): New.
32634         (split_address_groups): Check offset validity with above function.
32635         (gt-tree-ssa-loop-ivopts.h): Include header file.
32637 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32639         * config.gcc (arm*-*-*): Add missing 'fi'.
32641 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32643         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32644         included in -fopt-info does not matter.
32645         * doc/optinfo.texi (-fopt-info): Fix description of default
32646         behavour. Explicitly say order of options included in -fopt-info
32647         does not matter.
32649 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32651         * config.gcc: Allow combinations of aprofile and rmprofile values for
32652         --with-multilib-list.
32653         * config/arm/t-multilib: New file.
32654         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32655         variables.  Remove setting of ISA and floating-point ABI in
32656         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32657         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32658         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32659         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32660         CPU options.
32661         * config/arm/t-rmprofile: Likewise except for the matches changes.
32662         * doc/install.texi (--with-multilib-list): Document the combination of
32663         aprofile and rmprofile values and warn about pitfalls in doing that.
32665 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32667         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32668         (movdi_aarch64): Likewise.
32670 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32672         PR tree-optimization/80632
32673         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32674         field.
32675         (build_arrays): Initialize it for virtual phis.
32676         (fix_phi_nodes): Use it for virtual phis.
32678         PR tree-optimization/80558
32679         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32680         [x, y] op z into [x op, y op z] for op & or | if conditions
32681         are met.
32683 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32684             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32686         PR target/71607
32687         * config/arm/arm.md (use_literal_pool): Remove.
32688         (64-bit immediate split): No longer takes cost into consideration
32689         if arm_disable_literal_pool is enabled.
32690         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32691         used when arm_disable_literal_pool is enabled.
32692         (arm_max_const_double_inline_cost): Remove use of
32693         arm_disable_literal_pool.
32694         (push_minipool_fix): Add assert.
32695         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32696         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32697         (no_literal_pool_sf_immediate): New.
32699 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32701         PR tree-optimization/80613
32702         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32703         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32705 2017-05-05  Richard Biener  <rguenther@suse.de>
32707         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32709 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32711         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32712         of this flag from insn conditions due to removal from r247495.
32714 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32716         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32717         New function.
32718         (arm_early_store_addr_dep_ptr): Likewise.
32719         * config/arm/aarch-common-protos.h
32720         (arm_early_load_addr_dep_ptr): Add prototype.
32721         (arm_early_store_addr_dep_ptr): Likewise.
32722         * config/arm/cortex-a53.md: Add new bypasses.
32724 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32726         * tree.c (next_type_uid): Change type to unsigned.
32727         (type_hash_canon): Decrement back next_type_uid if
32728         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32729         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32730         if possible.
32732 2017-05-04  Martin Sebor  <msebor@redhat.com>
32734         * builtins.c: Fix a trivial typo in a comment.
32736         PR middle-end/79234
32737         * builtins.c (check_sizes): Adjust to handle reading past the end.
32738         Avoid printing excessive upper bound of ranges.  Use %E to print
32739         tree nodes instead of converting them to %wu.
32740         (expand_builtin_memchr): New function.
32741         (compute_dest_size): Rename...
32742         (compute_objsize): ...to this.
32743         (expand_builtin_memcpy): Adjust.
32744         (expand_builtin_mempcpy): Adjust.
32745         (expand_builtin_strcat): Adjust.
32746         (expand_builtin_strcpy): Adjust.
32747         (check_strncat_sizes): Adjust.
32748         (expand_builtin_strncat): Adjust.
32749         (expand_builtin_strncpy): Adjust and simplify.
32750         (expand_builtin_memset): Adjust.
32751         (expand_builtin_bzero): Adjust.
32752         (expand_builtin_memcmp): Adjust.
32753         (expand_builtin): Handle memcmp.
32754         (maybe_emit_chk_warning): Check strncat just once.
32756 2017-05-04  Martin Sebor  <msebor@redhat.com>
32758         PR preprocessor/79214
32759         PR middle-end/79222
32760         PR middle-end/79223
32761         * builtins.c (check_sizes): Add inlining context and issue
32762         warnings even when -Wno-system-headers is set.
32763         (check_strncat_sizes): Same.
32764         (expand_builtin_strncat): Same.
32765         (expand_builtin_memmove): New function.
32766         (expand_builtin_stpncpy): Same.
32767         (expand_builtin): Handle memmove and stpncpy.
32769 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32771         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32772         which is not used any more.
32774 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32776         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32778 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32780         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32781         (cortexa53_tunings): Likewise.
32782         (cortexa57_tunings): Likewise.
32783         (cortexa72_tunings): Likewise.
32784         (cortexa73_tunings): Likewise.
32786 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32788         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32789         Set loop alignment to 8.
32791 2017-05-04  Martin Sebor  <msebor@redhat.com>
32793         PR translation/80280
32794         * builtins.c (expand_builtin_object_size): Add missing quoting to
32795         %D and like directives.
32796         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32797         (hsa_type_for_tree_type): Same.
32798         (verify_function_arguments): Same.
32799         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32800         * varasm.c (get_section): Same.
32801         (mark_weak): Same.
32803 2017-05-04  Martin Sebor  <msebor@redhat.com>
32805         PR translation/80280
32806         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32808 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32810         * config/aarch64/aarch64.c (generic_addrcost_table):
32811         Change HI/TI mode setting.
32813 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32815         PR tree-optimization/80622
32816         * tree-sra.c (comes_initialized_p): New function.
32817         (build_accesses_from_assign): Only set write lazily when
32818         comes_initialized_p is false.
32819         (analyze_access_subtree): Use comes_initialized_p.
32820         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32821         instead of testing for PARM_DECL.
32823 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32825         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32826         constraint on operand 0 to allow more general addressing modes.
32827         Adjust output template.
32828         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32829         New function.
32830         * config/aarch64/aarch64-protos.h
32831         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32832         * config/aarch64/constraints.md (Dp): New address constraint.
32833         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32834         predicate.
32836 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32838         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32839         update use of estimate_ipcp_clone_size_and_time.
32840         (estimate_local_effects): Update use of
32841         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32842         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32843         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32844         Return nonspecialized time.
32846 2017-05-04  Richard Biener  <rguenther@suse.de>
32848         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32849         for the last VUSE which def dominates the PHI.  Directly call
32850         maybe_skip_until.
32851         (get_continuation_for_phi_1): Remove.
32853 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
32855         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32856         to explain the use of truncating division.  Cap the number of
32857         iterations to the maximum given by nb_iterations_upper_bound,
32858         if defined.
32860 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32862         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
32863         * configure: Regenerate.
32864         * config.in: Regenerate.
32865         * config/i386/driver-mingw32.c: new file.
32866         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
32867         * config.host: Link driver-mingw32.o on MinGW host.
32868         * doc/install.texi: Document new --enable-mingw-wildcard configure
32869         option.
32871 2017-05-04  Marek Polacek  <polacek@redhat.com>
32873         PR tree-optimization/80612
32874         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
32876 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32877             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
32879         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
32880         (movt splitter): Likewise.
32881         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
32882         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
32883         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
32884         block for Thumb-1 with MOVT.
32885         (thumb2_legitimate_address_p): Move code block ...
32886         (can_avoid_literal_pool_for_label_p): ... into this new function.
32887         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
32888         literal pool.
32889         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
32890         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
32891         "M-profile targets with the MOVT instruction".
32893 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32895         * config/arm/arm-builtins.c (arm_init_builtins): Rename
32896         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
32897         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
32899 2017-05-04  Martin Liska  <mliska@suse.cz>
32901         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
32902         variable cond_code.
32904 2017-05-04  Richard Biener  <rguenther@suse.de>
32906         * tree.c (array_at_struct_end_p): Handle arrays at struct
32907         end with flexarrays more conservatively.  Refactor and treat
32908         arrays of arrays or aggregates more strict.  Fix
32909         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
32910         * tree.c (array_at_struct_end_p): Adjust prototype.
32911         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
32912         * gimple-fold.c (get_range_strlen): Likewise.
32913         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
32915 2017-05-04  Richard Biener  <rguenther@suse.de>
32917         PR tree-optimization/31130
32918         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
32919         false.
32920         (supports_overflow_infinity): Likewise.
32921         (is_negative_overflow_infinity): Likewise.
32922         (is_positive_overflow_infinity): Likewise.
32923         (is_overflow_infinity): Likewise.
32924         (stmt_overflow_infinity): Likewise.
32925         (overflow_infinity_range_p): Likewise.
32926         (usable_range_p): Remove as always returning true.
32927         (make_overflow_infinity): Remove.
32928         (negative_overflow_infinity): Likewise.
32929         (positive_overflow_infinity): Likewise.
32930         (avoid_overflow_infinity): Likewise.
32931         (set_value_range): Adjust accordingly.
32932         (set_value_range_to_nonnegative): Likewise, remove now unused
32933         overflow_infinity arg.
32934         (vrp_operand_equal_p): Adjust.
32935         (update_value_range): Likewise.
32936         (range_int_cst_singleton_p): Likewise.
32937         (operand_less_p): Likewise.
32938         (compare_values_warnv): Likewise.
32939         (extract_range_for_var_from_comparison_expr): Likewise.
32940         (vrp_int_const_binop): Likewise.
32941         (zero_nonzero_bits_from_vr): Likewise.
32942         (extract_range_from_multiplicative_op_1): Likewise.
32943         (extract_range_from_binary_expr_1): Likewise.
32944         (extract_range_from_unary_expr): Likewise.
32945         (extract_range_from_comparison): Likewise.
32946         (extract_range_basic): Likewise.
32947         (adjust_range_with_scev): Likewise.
32948         (compare_ranges): Likewise.
32949         (compare_range_with_value): Likewise.
32950         (dump_value_range): Likewise.
32951         (test_for_singularity): Likewise, remove strict_overflow_p parameter
32952         never used.
32953         (simplify_cond_using_ranges): Adjust.
32955 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
32957         * brig-builtins.def: Added a builtin for class_f64.
32958         * builtin-types.def: Added a builtin type needed by class_f64.
32960 2017-05-03  Jason Merrill  <jason@redhat.com>
32962         * timevar.def: Add TV_CONSTEXPR.
32964 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
32966         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
32968 2017-05-03  Martin Jambor  <mjambor@suse.cz>
32970         * ipa-prop.c (ipa_update_after_lto_read): Removed.
32971         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
32972         * ipa-cp.c (ipcp_propagate_stage): Do not call
32973         ipa_update_after_lto_read.
32974         * ipa-inline.c (ipa_inline): Likewise.
32976 2017-05-03  Martin Jambor  <mjambor@suse.cz>
32978         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
32979         tag.  Added a default constructor and a destructor.
32980         (ipa_edge_args_sum_t): New class;
32981         (ipa_edge_args_sum): Declare.
32982         (ipa_edge_args_vector): Remove declaration.
32983         (IPA_EDGE_REF): Use ipa_edge_args_sum.
32984         (ipa_free_edge_args_substructures): Remove declaration.
32985         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
32986         (ipa_edge_args_info_available_for_edge_p): Likewise.
32987         * ipa-prop.c (ipa_edge_args_vector): Removed.
32988         (edge_removal_hook_holder): Likewise.
32989         (edge_duplication_hook_holder): Likewise.
32990         (ipa_edge_args_sum): New variable.
32991         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
32992         ipa_edge_args_vector.
32993         (ipa_free_edge_args_substructures): Likewise.
32994         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
32995         ipa_edge_args_vector.
32996         (ipa_edge_removal_hook): Turned into method
32997         ipa_edge_args_sum_t::remove.
32998         (ipa_edge_duplication_hook): Turned into method
32999         ipa_edge_args_sum_t::duplicate.
33000         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33001         registering edge hooks.
33002         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33003         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33004         ipa_edge_args_sum instead of ipa_edge_args_vector.
33005         * ipa-profile.c (ipa_profile): Likewise.
33007 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33009         * symbol-summary.h (function_summary): New method exists.
33010         (function_summary::symtab_removal): Deallocate through release.
33011         (call_summary): New class.
33012         (gt_ggc_mx): New overload.
33013         (gt_pch_nx): Likewise.
33014         (gt_pch_nx): Likewise.
33016 2017-05-03  Jeff Law  <law@redhat.com>
33018         PR tree-optimization/78496
33019         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33020         from simplify_cond_using_ranges.  Split off code to walk
33021         backwards through casts into ...
33022         (simplify_cond_using_ranges_2): New function.
33023         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33024         (execute_vrp): After identifying jump threads, call
33025         simplify_cond_using_ranges_2.
33027 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33029         PR bootstrap/80609
33030         * ipa-inline.h (inline_summary): Add ctor.
33031         (create_ggc): Do not use ggc_cleared_alloc.
33033 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33034             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33036         * gcc.c (handle_braces): Support escaping in switch matching
33037         text.
33038         * doc/invoke.texi (Spec Files): Document it.
33039         Remove superfluous @code markup in items.
33041 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33043         * diagnostic-show-locus.c (struct column_range): New struct.
33044         (get_affected_columns): New function.
33045         (get_printed_columns): New function.
33046         (struct correction): New struct.
33047         (correction::ensure_capacity): New function.
33048         (correction::ensure_terminated): New function.
33049         (struct line_corrections): New struct.
33050         (line_corrections::~line_corrections): New dtor.
33051         (line_corrections::add_hint): New function.
33052         (layout::print_trailing_fixits): Reimplement in terms of the new
33053         classes.
33054         (selftest::test_overlapped_fixit_printing): New function.
33055         (selftest::diagnostic_show_locus_c_tests): Call it.
33057 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33059         Canonicalize canonical type hashing
33060         * tree.h (type_hash_canon_hash): Declare.
33061         * tree.c (type_hash_list, attribute_hash_list): Move into
33062         type_hash_canon_hash.
33063         (build_type_attribute_qual_variant): Break out hash code calc into
33064         type_hash_canon_hash.
33065         (type_hash_canon_hash): New.  Generic type hash computation.
33066         (build_range_type_1, build_array_type_1, build_function_type,
33067         build_method_type_directly, build_offset_type, build_complex_type,
33068         make_vector_type): Call it.
33070 2017-05-03  Richard Biener  <rguenther@suse.de>
33072         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33073         When all DRs have unknown misaligned do not always peel
33074         when there is a store but apply the same costing model as if
33075         there were only loads.
33077 2017-05-03  Richard Biener  <rguenther@suse.de>
33079         Revert
33080         PR tree-optimization/80492
33081         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33082         compare_base_decls returning dont-know properly.
33084 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33086         * config/arm/iterators.md (CCSI): New mode iterator.
33087         (arch): New mode attribute.
33088         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33089         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33090         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33091         code iterator for success result mode.
33092         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33093         the corresponding new insn generators.
33095 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33097         Revert r247509
33098         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33099         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33101 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33103         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33104         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33105         (DDR_A): Wrap DDR argument in brackets.
33106         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33107         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33108         (DDR_REVERSED_P): Likewise.
33110 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33112         PR tree-optimization/79472
33113         * tree-switch-conversion.c (struct switch_conv_info): Add
33114         contiguous_range and default_case_nonstandard fields.
33115         (collect_switch_conv_info): Compute contiguous_range and
33116         default_case_nonstandard fields, don't clear final_bb if
33117         contiguous_range and only the default case doesn't have the required
33118         structure.
33119         (check_all_empty_except_final): Set default_case_nonstandard instead
33120         of failing if contiguous_range and the default case doesn't have empty
33121         block.
33122         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33123         and only the default case doesn't have the required constants.  Skip
33124         virtual phis.
33125         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33126         if default_case_nonstandard.
33127         (build_constructors): Build constant 1 just once.  Assert that default
33128         values aren't inserted in between cases if contiguous_range.  Skip
33129         virtual phis.
33130         (build_arrays): Skip virtual phis.
33131         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33132         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33133         Handle virtual phis.
33134         (gen_inbound_check): Handle default_case_nonstandard case.
33135         (process_switch): Adjust check_final_bb caller.  Call
33136         gather_default_values with the first non-default case instead of
33137         default case if default_case_nonstandard.
33139 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33141         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33142         check.  Fix formatting.
33144 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33146         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33147         errors when comparing specialized and unspecialized times.
33149 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33151         * diagnostic-show-locus.c
33152         (layout::should_print_annotation_line_p): Make private.
33153         (layout::print_annotation_line): Make private.
33154         (layout::annotation_line_showed_range_p): Make private.
33155         (layout::show_ruler): Make private.
33156         (layout::print_source_line): Make private.  Pass in line and
33157         line_width, rather than calling location_get_source_line.  Drop
33158         returned value.
33159         (layout::print_leading_fixits): New method.
33160         (layout::print_any_fixits): Rename to...
33161         (layout::print_trailing_fixits): ...this, and make private.
33162         Don't print newline fixits.
33163         (diagnostic_show_locus): Move logic for printing one row into...
33164         (layout::print_line): ...this new function.  Move the
33165         location_get_source_line call and error-handling from
33166         print_source_line to here.  Call print_leading_fixits, and rename
33167         print_any_fixits to print_trailing_fixits.
33168         (selftest::test_fixit_insert_containing_newline): Update now that
33169         newlines are partially supported.
33170         (selftest::test_fixit_insert_containing_newline_2): New test.
33171         (selftest::test_fixit_replace_containing_newline): Update comments.
33172         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33173         * edit-context.c (class added_line): New class.
33174         (class edited_line): Describe newline handling in comment.
33175         (edited_line::actually_edited_p): New method.
33176         (edited_line::print_content): Delete redundant decl.
33177         (edited_line::m_predecessors): New field.
33178         (edited_file::print_content): Call edited_line::print_content.
33179         (edited_file::print_diff): Update to support newlines.
33180         (edited_file::print_diff_hunk): Likewise.
33181         (edited_file::print_run_of_changed_lines): New function.
33182         (edited_file::print_diff_line): Convert to...
33183         (print_diff_line): ...this.
33184         (edited_file::get_effective_line_count): New function.
33185         (edited_line::edited_line): Initialize new field m_predecessors.
33186         (edited_line::~edited_line): Clean up m_predecessors.
33187         (edited_line::apply_fixit): Handle newlines.
33188         (edited_line::get_effective_line_count): New function.
33189         (edited_line::print_content): New function.
33190         (edited_line::print_diff_lines): New function.
33191         (selftest::test_applying_fixits_insert_containing_newline): New
33192         test.
33193         (selftest::test_applying_fixits_replace_containing_newline): New
33194         test.
33195         (selftest::insert_line): New function.
33196         (selftest::test_applying_fixits_multiple_lines): Add example of
33197         inserting a line.
33198         (selftest::edit_context_c_tests): Call the new tests.
33200 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33202         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33203         parameter cand.  Update dump information.
33204         (get_computation_cost): Update uses.
33206 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33208         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33209         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33210         (get_computation_at, rewrite_use_address): Update use of
33211         get_computation_aff.
33213 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33215         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33216         (get_computation): Delete.
33217         (get_computation_cost): Implement like get_computation_cost_at.
33218         Use get_computation_at.
33219         (get_computation_cost_at): Delete.
33220         (rewrite_use_nonlinear_expr): Use get_computation_at.
33221         (rewrite_use_compare, remove_unused_ivs): Ditto.
33223 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33225         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33227 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33229         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33230         (ivopts_global_cost_for_size): Rename parameter and update uses.
33231         (iv_ca_recount_cost): Update uses.
33232         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33233         candidates seperately in n_invs and n_cands.
33234         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33236 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33238         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33239         (find_inv_vars_cb): New.
33240         (find_depends): Renamed to ...
33241         (find_inv_vars): ... this.
33242         (add_candidate_1, force_var_cost): Call find_inv_vars.
33243         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33245 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33247         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33248         inv_vars.  Add inv_exprs.
33249         (struct iv_cand): Rename depends_on to inv_vars.
33250         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33251         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33252         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33253         (dump_cand): Dump inv_vars.
33254         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33255         (record_invariant, find_depends, add_candidate_1): Ditto.
33256         (set_group_iv_cost, force_var_cost): Ditto.
33257         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33258         (get_computation_cost_at, get_computation_cost): Ditto.
33259         (determine_group_iv_cost_generic): Ditto.
33260         (determine_group_iv_cost_address): Ditto.
33261         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33262         (determine_group_iv_costs): Ditto.
33263         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33264         (iv_ca_set_remove_invariants): Renamed to ...
33265         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33266         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33267         (iv_ca_set_add_invariants):  Renamed to ...
33268         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33269         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33270         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33271         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33272         (create_new_ivs): Remove useless dump.
33274 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33276         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33277         iv_cand code.
33278         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33279         (iv_ca_set_no_cp, create_new_iv): Ditto.
33281 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33283         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33285 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33287         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33288         function tree_check2.
33290 2017-05-02  Martin Liska  <mliska@suse.cz>
33292         * doc/gcov.texi: Add missing preposition.
33293         * gcov.c (function_info::function_info): Properly fill up
33294         all member variables.
33296 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33298         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33300 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33302         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33304 2017-05-02  Martin Liska  <mliska@suse.cz>
33306         PR lto/77954.
33307         * lto-streamer-in.c (lto_read_tree_1): Remove
33308         LTO_STREAMER_DEBUG.
33309         * lto-streamer.c (struct tree_hash_entry): Likewise.
33310         (struct tree_entry_hasher): Likewise.
33311         (tree_entry_hasher::hash): Likewise.
33312         (tree_entry_hasher::equal): Likewise.
33313         (lto_streamer_init): Likewise.
33314         (lto_orig_address_map): Likewise.
33315         (lto_orig_address_get): Likewise.
33316         (lto_orig_address_remove): Likewise.
33317         * lto-streamer.h: Likewise.
33318         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33319         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33321 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33323         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33324         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33325         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33326         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33327         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33328         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33329         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33330         (mm_maskz_sub_ss): New intrinsics.
33331         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33332         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33333         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33334         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33335         (__builtin_ia32_subss_mask_round): New builtins.
33336         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33337         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33338         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33339         Renamed to ...
33340         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33341         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33342         Changed to ...
33343         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33344         ... this.
33346 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33348         PR tree-optimization/78687
33349         * tree-sra.c (access): New field parent.
33350         (process_subtree_disqualification): New function.
33351         (disqualify_candidate): Call it.
33352         (build_accesses_from_assign): Reset write flag if creating an
33353         assighnment link.
33354         (build_access_subtree): Fill in parent field and also prpagate
33355         down grp_write flag.
33356         (create_artificial_child_access): New parameter set_grp_write, set
33357         grp_write to its value.
33358         (propagate_subaccesses_across_link): Also propagate grp_write flag
33359         values.
33360         (propagate_all_subaccesses): Push the closest parent back to work
33361         queue if add_access_to_work_queue returned true.
33363 2017-05-02  Richard Biener  <rguenther@suse.de>
33365         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33366         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33367         -fstrict-overflow documentation.
33368         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33369         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33370         flag_strict_overflow.
33371         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33372         * lto-opts.c (lto_write_options): Do not stream it.
33373         * lto-wrapper.c (merge_and_complain): Do not handle it.
33374         * opts.c (default_options_table): Do not set -fstrict-overflow.
33375         (finish_options): Likewise do not clear it when sanitizing.
33376         * simplify-rtx.c (simplify_const_relational_operation): Do not
33377         test flag_strict_overflow.
33379 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33381         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33382         using enabled attribute.
33383         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33384         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33385         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33386         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33387         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33388         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33389         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33390         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33391         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33392         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33394 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33396         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33398 2017-05-02  Richard Biener  <rguenther@suse.de>
33400         PR tree-optimization/80591
33401         Revert
33402         2017-04-10  Richard Biener  <rguenther@suse.de>
33404         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33405         asm inputs.
33407 2017-05-02  Richard Biener  <rguenther@suse.de>
33409         PR tree-optimization/80549
33410         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33411         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33412         headers if they do not have a preheader.
33414 2017-05-02  Martin Liska  <mliska@suse.cz>
33416         PR other/80589
33417         * common.opt: Fix typo.
33418         * doc/invoke.texi: Likewise.
33420 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33422         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33423         swapping, add (x,x,m,x,n) alternative.
33425 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33427         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33428         unnecessary unadjusted_alignment check.
33430 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33432         PR c++/80038
33433         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33434         operations here.
33435         * gimplify.c (gimplify_cilk_detach): New function.
33436         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33437         * tree-core.h: Document EXPR_CILK_SPAWN.
33438         * tree.h (EXPR_CILK_SPAWN): Define.
33440 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33442         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33443         to use new fixit_hint representation, using the "replace" logic.
33444         (get_line_span_for_fixit_hint): Likewise.
33445         (layout::print_any_fixits): Likewise.
33446         (selftest::test_one_liner_many_fixits): Rename to...
33447         (selftest::test_one_liner_many_fixits_1): ...this, and update
33448         comment and expected output to reflect that the multiple fix-it
33449         hints are now consolidated into one insertion.
33450         (selftest::test_one_liner_many_fixits_2): New test.
33451         (selftest::test_diagnostic_show_locus_one_liner): Update for
33452         above.
33453         (selftest::test_fixit_consolidation): Update for fix-it API
33454         change.
33455         * diagnostic.c (print_parseable_fixits): Likewise.
33456         * edit-context.c (edited_line::m_line_events): Convert from
33457         auto_vec <line_event *> to auto_vec <line_event>.
33458         (class line_event): Convert from abstract base class to a concrete
33459         class, taking over the role of replace_event.
33460         (class insert_event): Delete.
33461         (class replace_event): Rename to class line_event.  Convert to
33462         half-open range.
33463         (edit_context::add_fixits): Reimplement.
33464         (edit_context::apply_insert): Delete.
33465         (edit_context::apply_replace): Rename to...
33466         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33467         (edited_file::apply_insert): Delete.
33468         (edited_file::apply_replace): Rename to...
33469         (edited_file::apply_fixit): ...this.
33470         (edited_line::~edited_line): Drop deletion of events.
33471         (edited_line::apply_insert): Delete.
33472         (edited_line::apply_replace): Rename to...
33473         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33474         Update for change to type of m_line_events.
33475         * edit-context.h (edit_context::apply_insert): Delete.
33476         (edit_context::apply_replace): Rename to...
33477         (edit_context::apply_fixit): ...this.
33479 2017-05-01  Martin Sebor  <msebor@redhat.com>
33481         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33482         known.
33484 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33486         PR target/68491
33487         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33488         __get_cpuid_max returns 0.
33489         (__get_cpuid_count): Ditto.
33491 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33493         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33494         replacement expression is another instance of one of its arguments.
33496 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33498         PR target/79430
33499         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33500         check for stack push/pop autoinc.
33501         * config/i386/i386.c (ix86_agi_dependent): Return false
33502         if the only reason why modified_in_p returned true is that
33503         addr is SP based and set_insn is a push or pop.
33505 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33507         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33508         overflow check.
33510 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33512         PR ipa/79224
33513         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33514         (account_size_time): Use two predicates - exec_pred and
33515         nonconst_pred_ptr.
33516         (evaluate_conditions_for_known_args): Compute both clause and
33517         nonspec_clause.
33518         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33519         (inline_summary_t::duplicate): Update.
33520         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33521         separately.
33522         (compute_inline_parameters): Likewise.
33523         (estimate_edge_size_and_time): Update caluclation of time.
33524         (estimate_node_size_and_time): Compute both time and nonspecialized
33525         time.
33526         (estimate_ipcp_clone_size_and_time): Update.
33527         (inline_merge_summary): Update.
33528         (do_estimate_edge_time): Update.
33529         (do_estimate_edge_size): Update.
33530         (do_estimate_edge_hints): Update.
33531         (inline_read_section, inline_write_summary): Stream both new predicates.
33532         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33533         as argument.
33534         (compute_inlined_call_time): Cleanup.
33535         (big_speedup_p): Update.
33536         (edge_badness): Update.
33537         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33538         (size_time_entry): Replace predicate by exec_predicate and
33539         nonconst_predicate.
33540         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33541         (estimate_edge_time): Return also nonspec_time.
33542         (reset_edge_growth_cache): Update.
33544 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33546         PR rtl-optimization/80491
33547         * ifcvt.c (noce_process_if_block): When looking for x setter
33548         with missing else_bb, don't check only the insn right before
33549         cond_earliest, but look for the last insn that x is modified in
33550         within the same bb.
33552         PR rtl-optimization/80491
33553         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33555 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33557         PR tree-optimization/80487
33558         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33560 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33562         PR tree-optimization/79697
33563         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33564         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33565         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33566         BUILT_IN_STRNDUP.
33567         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33568         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33570 2017-04-28  Martin Sebor  <msebor@redhat.com>
33572         PR tree-optimization/80523
33573         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33574         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33575         functions.
33576         (maybe_warn, format_directive, parse_directive): Use new functions.
33577         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33579 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33581         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33583 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33585         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33586         target_header_dir): Set correctly.
33587         * configure: Regenerated.
33588         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33589         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33590         instead of SYSTEM_HEADER_DIR.
33592 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33594         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33595         (estimate_local_effects): Likewise.
33596         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33597         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33598         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33599         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33600         do_estimate_edge_time, estimate_edge_time): Likewise.
33601         * ipa-inline-analysis.c (estimate_node_size_and_time,
33602         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33603         (estimate_time_after_inlining): Remove.
33605 2017-04-28  Martin Liska  <mliska@suse.cz>
33607         * doc/gcov.texi: Enhance documentation of gcov.
33609 2017-04-28  Martin Liska  <mliska@suse.cz>
33611         * doc/gcov.texi: Sort options in alphabetic order.
33612         * doc/gcov-dump.texi: Likewise.
33613         * doc/gcov-tool.texi: Likewise.
33614         * gcov.c (print_usage): Likewise.
33615         * gcov-dump.c (print_usage): Likewise.
33616         * gcov-tool.c (print_merge_usage_message): Likewise.
33617         (print_rewrite_usage_message): Likewise.
33618         (print_overlap_usage_message): Likewise.
33620 2017-04-28  Martin Liska  <mliska@suse.cz>
33622         PR gcov-profile/53915
33623         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33625 2017-04-28  Martin Liska  <mliska@suse.cz>
33627         PR gcov-profile/79891
33628         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33629         is marked by compiler as living on a line.
33630         (get_cycles_count): Remove usage of the union.
33631         (output_intermediate_file): Likewise.
33632         (find_source): Fix GNU coding style.
33633         (accumulate_line_counts): Remove old non-all block mode.
33634         (output_lines): Remove usage of the union.
33635         * profile.c (output_location): Include all BBs, even if
33636         belonging to a same line (and file) as a previous BB.
33638 2017-04-28  Martin Liska  <mliska@suse.cz>
33640         * gcov.c (process_args): Handle new argument 'w'.
33641         (read_graph_file): Assign ID to BBs.
33642         (output_branch_count): Display BB # if verbose flag is set.
33643         (output_lines): Likewise for arcs.
33644         (print_usage): Add '--verbose' option help.
33645         * doc/gcov.texi: Document --verbose (-w) option.
33647 2017-04-28  Martin Liska  <mliska@suse.cz>
33649         * gcov.c (struct block_location_info): New struct.
33650         (process_file): Fill up the new structure.
33651         (read_graph_file): Replace usage of encoding by the newly added
33652         struct.
33653         (add_line_counts): Likewise.
33654         (accumulate_line_counts): Remove usage of the union.
33655         (function_info::function_info): New function.
33656         (function_info::~function_info): Likewise.
33657         (process_file): Call delete instead of release_function.
33658         (release_function): Release the function.
33659         (release_structures): Call delete instead of release_function.
33660         (solve_flow_graph): Replace usage of num_blocks.
33661         (find_exception_blocks): Likewise.
33662         (output_lines): Fix GNU coding style.
33664 2017-04-28  Martin Liska  <mliska@suse.cz>
33666         PR driver/56469
33667         * coverage.c (coverage_remove_note_file): New function.
33668         * coverage.h: Declare the function.
33669         * toplev.c (finalize): Clean if an error has been seen.
33671 2017-04-28  Martin Liska  <mliska@suse.cz>
33673         PR gcov-profile/80031
33674         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33675         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33676         * gcov.c (read_graph_file): Read just number of blocks.
33677         * profile.c (branch_prob): Do not stream 0 flags per a basic
33678         block.
33680 2017-04-28  Martin Liska  <mliska@suse.cz>
33682         * gcov-dump.c (tag_*): Add new argument to declarations.
33683         (dump_gcov_file): Likewise.
33684         (tag_blocks): Add and use new argument depth.
33685         (tag_arcs): Likewise.
33686         (tag_lines): Likewise.
33687         (tag_counters): Likewise.
33688         (tag_summary): Likewise.
33689         (dump_working_sets): Use depth to do a proper indentation.
33691 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33693         PR bootstrap/80531
33694         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33695         * symtab.c (symtab_node::debug_symtab): Move definition here.
33697 2017-04-28  Richard Biener  <rguenther@suse.de>
33699         * lto-streamer.h (LTO_major_version): Bump to 7.
33701 2017-04-28  Richard Biener  <rguenther@suse.de>
33703         * tree-vrp.c (assert_info): New struct.
33704         (add_assert_info): New helper.
33705         (register_edge_assert_for_2): Refactor to add asserts to a vector
33706         of assert_info.
33707         (register_edge_assert_for_1): Likewise.
33708         (register_edge_assert_for): Likewise.
33709         (finish_register_edge_assert_for): New helper actually registering
33710         asserts where live on edge.
33711         (find_conditional_asserts): Adjust.
33712         (find_switch_asserts): Likewise.
33713         (evrp_dom_walker::try_find_new_range): Generalize.
33714         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33716 2017-04-27  Marek Polacek  <polacek@redhat.com>
33718         PR sanitizer/80349
33719         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33720         arg10 and arg11 to itype.
33722 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33724         * doc/extend.texi (Object Size Checking): Improve grammar.
33726 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33728         PR target/80530
33729         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33730         that the logic for permitting reciprocal estimates matches that
33731         in use_rsqrt_p.
33733 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33735         PR c++/80534
33736         * tree.c (type_cache_hasher::equal): Only compare
33737         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33738         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33739         non-aggregate element types.
33740         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33741         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33743 2017-04-27  Richard Biener  <rguenther@suse.de>
33745         PR middle-end/80533
33746         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33747         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33748         keeping a reference to a trailing array.
33750 2017-04-27  Richard Biener  <rguenther@suse.de>
33752         PR middle-end/80539
33753         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33754         being in loop-closed SSA form conservatively.
33755         (chrec_fold_multiply_poly_poly): Likewise.
33757 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33759         PR middle-end/79665
33760         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33761         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33763 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33765         PR target/77728
33766         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33767         (aarch64_function_arg_alignment): Return unsigned int again, but still
33768         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33769         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33770         Don't emit -Wpsabi note.
33771         (aarch64_function_arg_boundary): Likewise.
33772         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33773         caller.
33775 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33777         * tree.h (crc32_unsigned_n): Declare.
33778         (crc32_unsigned, crc32_unsigned): Make inline.
33779         * tree.c (crc32_unsigned_bits): Replace with ...
33780         (crc32_unsigned_n): ... this.
33781         (crc32_unsigned, crc32_byte): Remove.
33782         (crc32_string): Remove unnecessary braces.
33784 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33786         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33787         * ipa-inline-analysis.c (MAX_TIME): Remove.
33788         (account_size_time): Use sreal for time.
33789         (dump_inline_summary): Update.
33790         (estimate_function_body_sizes): Update.
33791         (estimate_edge_size_and_time): Update.
33792         (estimate_calls_size_and_time): Update.
33793         (estimate_node_size_and_time): Update.
33794         (inline_merge_summary): Update.
33795         (inline_update_overall_summary): Update.
33796         (estimate_time_after_inlining): Update.
33797         (inline_read_section): Update.
33798         (inline_write_summary): Update.
33799         * ipa-inline.c (compute_uninlined_call_time): Update.
33800         (compute_inlined_call_time): Update.
33801         (recursive_inlining): Update.
33802         (inline_small_functions): Update.
33803         (dump_overall_stats): Update.
33804         * ipa-inline.h: Include sreal.h.
33805         (size_time_entry): Turn time to sreal.
33806         (inline_summary): Turn self_time nad time to sreal.
33808 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33810         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33811         data-streamer.h
33812         (sreal::stream_out, sreal::stream_in): New.
33813         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33815 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33817         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33818         environment.
33820 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33822         PR target/70799
33823         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33824         Handle ASHIFTRT.
33825         (dimode_scalar_chain::compute_convert_gain): Ditto.
33826         (dimode_scalar_chain::make_vector_copies): Ditto.
33827         (dimode_scalar_chain::convert_reg): Ditto.
33828         (dimode_scalar_chain::convert_insn): Ditto.
33829         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33830         (VI248_AVX512BW_1): New mode iterator.
33831         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33832         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33833         mode iterator.
33835 2017-04-25  Martin Sebor  <msebor@redhat.com>
33837         PR tree-optimization/80497
33838         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33839         constants are representable in HOST_WIDE_INT.
33840         (parse_directive): Ditto.
33842 2017-04-25  Martin Sebor  <msebor@redhat.com>
33844         PR bootstrap/80486
33845         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33846         (new_zero_array): Adjust signature.
33847         (dom_info::dom_init): Used unsigned rather that size_t.
33848         (dom_info::dom_info): Same.
33850 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33851             Jakub Jelinek  <jakub@redhat.com>
33853         PR target/77728
33854         * config/arm/arm.c: Include gimple.h.
33855         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33856         returns negative, increment ncrn only if it returned positive.
33857         (arm_needs_doubleword_align): Return int instead of bool,
33858         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33859         members, but if there is any such non-FIELD_DECL
33860         > PARM_BOUNDARY aligned decl, return -1 instead of false.
33861         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33862         returns negative, increment nregs only if it returned positive.
33863         (arm_setup_incoming_varargs): Likewise.
33864         (arm_function_arg_boundary): Emit -Wpsabi note if
33865         arm_needs_doubleword_align returns negative, return
33866         DOUBLEWORD_ALIGNMENT only if it returned positive.
33868 2017-04-25  Marek Polacek  <polacek@redhat.com>
33870         PR sanitizer/80349
33871         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
33872         first argument to type.
33874 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
33876         PR target/80482
33877         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
33878         type checks to test for compatibility instead of equality.
33880 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33881             Jakub Jelinek  <jakub@redhat.com>
33883         PR target/77728
33884         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
33885         type.
33886         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
33887         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
33888         the alignment computation, but return their maximum in warn_alignment.
33889         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33890         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
33891         is smaller.
33892         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
33893         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33894         caller.
33896 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33898         * config/arc/simdext.md (dmpyh): Fix typo.
33900 2017-04-25  Richard Biener  <rguenther@suse.de>
33902         PR tree-optimization/80492
33903         * alias.c (compare_base_decls): Handle registers with asm
33904         specification conservatively.
33905         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33906         compare_base_decls returning dont-know properly.
33908 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33910         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
33911         (legitimate_offset_address_p): New function.
33912         (arc_legitimate_address_p): Use above function.
33914 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33916         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
33918 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33920         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
33921         ACCH registers whenever they are available.
33923 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33925         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
33926         double regs fix when not used.
33928 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33930         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
33931         core registers.
33932         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
33933         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
33935 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33937         * config/arc/arc.c (arc_output_addsi): Check for h-register class
33938         when emitting short ADD instructions.
33940 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33942         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
33943         constraint.
33944         (cmpsi_cc_c_insn): Likewise.
33945         (cbranchsi4_scratch): Compute proper instruction length using
33946         compact_hreg_operand.
33947         * config/arc/predicates.md (compact_hreg_operand): New predicate.
33949 2017-04-25  Richard Biener  <rguenther@suse.de>
33951         PR middle-end/80509
33952         * passes.c (pass_manager::pass_manager): Initialize
33953         m_name_to_pass_map.
33955 2017-04-25  Richard Biener  <rguenther@suse.de>
33957         PR tree-optimization/79201
33958         * tree-ssa-sink.c (statement_sink_location): Handle calls.
33960 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
33962         PR target/80464
33963         * config/s390/vector.md: Split MEM->GPR vector moves for
33964         non-s_operand addresses.
33966 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
33968         PR target/79895
33969         * config/s390/predicates.md (reload_const_wide_int_operand): New
33970         predicate.
33971         * config/s390/s390.md ("movti"): Remove d/P alternative.
33972         ("movti_bigconst"): New pattern definition.
33974 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
33976         PR target/80080
33977         * s390-protos.h (s390_expand_cs_hqi): Removed.
33978         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
33979         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
33980         modes as well as CCZ1mode and CCZmode.
33981         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
33982         signature of s390_emit_compare_and_swap.
33983         (s390_expand_cs_hqi): Likewise, make static.
33984         (s390_expand_cs_tdsi): Generate an explicit compare before trying
33985         compare-and-swap, in some cases.
33986         (s390_expand_cs): Wrapper function.
33987         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
33988         atomic_exchange.
33989         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
33990         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
33991         patterns for small and large integers.  Forbid symref memory operands.
33992         Move expander to s390.c.  Require cc register.
33993         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
33994         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
33995         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
33996         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
33997         symref memory operands.  Remove CC mode and call s390_match_ccmode
33998         instead.
33999         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34001 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34003         * config/s390/s390.md (define_peephole2): New peephole to help
34004         combining the load-and-test pattern with volatile memory.
34006 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34008         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34009         with CCZmode for TARGET_Z196.
34011 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34013         PR rtl-optimization/80501
34014         * combine.c (make_compound_operation_int): Set subreg_code to SET
34015         even for AND with mask of the sign bit of mode.
34017         PR rtl-optimization/80500
34018         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34019         sum's initial value.
34021 2017-04-25  Julian Brown  <julian@codesourcery.com>
34022             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34024         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34026 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34028         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34030 2017-04-25  Julian Brown  <julian@codesourcery.com>
34031             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34033         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34034         (thunderx2t99_sha): New Reservation.
34036 2017-04-25  Julian Brown  <julian@codesourcery.com>
34037             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34039         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34040         type for 1-element load.
34042 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34044         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34046 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34048         PR tree-optimization/80293
34049         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34050         char arrays not totally scalarizable if it is false.
34051         (analyze_all_variable_accesses): Pass correct value in the new
34052         parameter.  Add a statistics counter.
34054 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34056         PR middle-end/79931
34057         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34059 2017-04-24  Richard Biener  <rguenther@suse.de>
34061         PR tree-optimization/80494
34062         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34063         out for complex types.
34065 2017-04-24  Richard Biener  <rguenther@suse.de>
34067         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34068         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34069         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34070         (DFS): Adjust and never fail.
34071         (sccvn_dom_walker::fail): Remove.
34072         (sccvn_dom_walker::before_dom_children): Adjust.
34073         (run_scc_vn): Likewise and never fail.
34074         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34075         (pass_fre::execute): Likewise.
34077 2017-04-24  Richard Biener  <rguenther@suse.de>
34079         PR tree-optimization/79725
34080         * tree-ssa-sink.c (statement_sink_location): Return whether
34081         failure reason was zero uses.  Move that check later.
34082         (sink_code_in_bb): Deal with zero uses by removing the stmt
34083         if possible.
34085 2017-04-24  Richard Biener  <rguenther@suse.de>
34087         PR c++/2972
34088         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34089         pointer-based references.
34091 2017-04-24  Richard Biener  <rguenther@suse.de>
34093         PR bootstrap/79814
34094         * pass_manager.h (pass_manager::operator new): Remove.
34095         (pass_manager::operator delete): Likewise.
34096         * passes.c (pass_manager::operator new): Remove.
34097         (pass_manager::operator delete): Likewise.
34098         (pass_manager::pass_manager): Zero individual pass members.
34100 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34102         PR target/70799
34103         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34104         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34105         Check "XEXP (src, 1)" operand here.
34106         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34107         Check "XEXP (src, 1)" operand here.
34108         (dimode_scalar_chain::make_vector_copies): Detect count register
34109         of a shift instruction.  Zero extend count register from QImode
34110         to DImode to satisfy vector shift pattern count operand predicate.
34111         Substitute vector shift count operand with a DImode copy.
34112         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34113         vector register.
34115 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34117         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34118         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34119         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34120         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34121         (UNSPEC_NOREX_MEM): Remove definition.
34123 2017-04-21  Richard Biener  <rguenther@suse.de>
34125         PR tree-optimization/79547
34126         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34127         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34128         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34129         without any constraints.
34131 2017-04-21  Richard Biener  <rguenther@suse.de>
34133         PR tree-optimization/78847
34134         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34136 2017-04-21  Richard Biener  <rguenther@suse.de>
34138         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34139         (build_distinct_type_copy): Likewise.
34140         (build_variant_type_copy): Likewise.
34141         * tree.c (build_qualified_type): Pass down mem-stat info.
34142         (build_distinct_type_copy): Likewise.
34143         (build_variant_type_copy): Likewise.
34145 2017-04-21  Richard Biener  <rguenther@suse.de>
34147         PR tree-optimization/80237
34148         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34149         defaulted to NULL.
34150         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34151         for a simplified result.
34153 2016-04-21  Richard Biener  <rguenther@suse.de>
34155         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34156         sth as strict as a simple_iv but a chrec without symbols and an
34157         operand defined in the loop we are peeling (and not some subloop).
34158         (propagate_constants_for_unrolling): Propagate all constants.
34160 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34162         PR target/79804
34163         * config/i386/i386.c (print_reg): Remove assert for disalowed
34164         regno values, call output_operand_lossage instead.
34166 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34168         PR target/78090
34169         * config/i386/constraints.md (Yc): New register constraint.
34170         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34171         Use Yc constraint for alternative 2 of operand 0.  Remove
34172         preferred_for_speed attribute.
34174 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34176         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34177         lastprivate clauses in SIMT case.
34179 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34181         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34183 2017-04-20  Richard Biener  <rguenther@suse.de>
34185         PR tree-optimization/57796
34186         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34187         as N scalar stores.
34188         (vect_model_load_cost): Cost gathers as N scalar loads.
34190 2017-04-20  Richard Biener  <rguenther@suse.de>
34192         * ggc-page.c (ggc_allocated_p): Rename to ...
34193         (safe_lookup_page_table_entry): ... this and return the lookup
34194         result.
34195         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34197 2017-04-20  Richard Biener  <rguenther@suse.de>
34199         PR tree-optimization/80453
34200         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34201         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34202         from the conditions.
34203         (vn_phi_eq): Pass them down.
34204         (vn_phi_lookup): Record them.
34205         (vn_phi_insert): Likewise.
34207 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34209         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34210         uninitialized variable warning to avoid buffer overrun.
34212 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34214         PR other/71250
34215         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34216         is suppressed for '{ 0 }' in C.
34218 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34220         * BASE-VER: Set to 8.0.0.
34222 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34224         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34225         priority .init_array and .fini_array section with SECTION_NOTYPE
34226         flag.
34228 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34230         PR middle-end/80423
34231         * tree.h (build_array_type): Add typeless_storage default argument.
34232         * tree.c (type_cache_hasher::equal): Also compare
34233         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34234         (build_array_type): Add typeless_storage argument, set
34235         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34236         recursive call.
34237         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34238         (build_array_type): Likewise.  Add typeless_storage argument.
34240 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34241             Jakub Jelinek  <jakub@redhat.com>
34243         PR tree-optimization/80426
34244         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34245         operation on symbolic operands, also compute the overflow for the
34246         invariant part when the operation degenerates into a negation.
34248 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34250         PR debug/80461
34251         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34252         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34254         PR debug/80436
34255         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34257 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34259         PR target/80462
34260         * config/avr/avr.c (tree.h): Include it.
34261         (cgraph.h): Include it.
34262         (avr_encode_section_info): Don't warn for uninitialized progmem
34263         variable if it's just an alias.
34265 2017-04-19  Richard Biener  <rguenther@suse.de>
34267         PR ipa/65972
34268         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34269         when needed by AutoPGO.
34271 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34273         PR lto/50345
34274         * doc/lto.texi: Remove an extra 'that'.
34276 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34278         PR rtl-optimization/80429
34279         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34280         are only used in debug insns.
34282 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34283             Vladimir Makarov  <vmakarov@redhat.com>
34285         * config/sparc/predicates.md (input_operand): Add comment.  Return
34286         true for any memory operand when LRA is in progress.
34287         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34289 2017-04-18  Jeff Law  <law@redhat.com>
34291         PR target/74563
34292         * mips.md ({return,simple_return}_internal): Do not overwrite
34293         operands[0].
34295 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34297         PR tree-optimization/80443
34298         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34299         instead of adding 1, subtract -1 and similarly instead of subtracting
34300         1 add -1.
34302 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34304         PR rtl-optimization/80357
34305         * haifa-sched.c (tmp_bitmap): New variable.
34306         (model_recompute): Handle duplicate use records.
34307         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34308         (free_global_sched_pressure_data): Free it.
34310 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34312         Revert:
34313         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34314         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34315         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34316         instead of SYSTEM_HEADER_DIR.
34318 2017-04-18  Jeff Law  <law@redhat.com>
34320         PR middle-end/80422
34321         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34322         predecessors after walking up the insn chain.
34324 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34326         PR debug/80263
34327         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34328         sizetype type into debug info.
34330 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34332         PR target/80099
34333         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34334         unneeded test for TARGET_UPPER_REGS_SF.
34335         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34337 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34339         PR sanitizer/80444
34340         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34341         instead of gsi_after_labels.
34343 2017-04-18  Jeff Law  <law@redhat.com>
34345         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34346         stack pointer.
34348         Revert:
34349         2017-04-13  Jeff Law  <law@redhat.com>
34350         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34351         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34353 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34355         PR target/79453
34356         * config/avr/avr.c (intl.h): Include it.
34357         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34359 2017-04-18  Martin Liska  <mliska@suse.cz>
34361         PR gcov-profile/78783
34362         * gcov-tool.c (gcov_output_files): Validate that destination
34363         file is either removed by the tool or by a user.
34365 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34366             Guy Benyei  <guybe@mellanox.com>
34368         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34369         block, and do not negate it, the stored id is already negative.
34371 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34373         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34375 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34377         PR target/80098
34378         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34379         masks of options that should be turned off if the VSX vector
34380         options are turned off.
34381         (OTHER_P8_VECTOR_MASKS): Likewise.
34382         (OTHER_VSX_VECTOR_MASKS): Likewise.
34383         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34384         rs6000_disable_incompatible_switches to validate no type switches
34385         like -mvsx.
34386         (rs6000_incompatible_switch): New function to disallow turning on
34387         other vector options if -mno-vsx, -mno-power8-vector, or
34388         -mno-power9-vector are specified.
34390 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34392         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34394 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34396         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34397         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34398         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34399         (ARG_POINTER_CFA_OFFSET): Likewise.
34401 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34403         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34404         conditions to take advantage of various optimizations.
34406 2017-04-13  Jeff Law  <law@redhat.com>
34408         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34409         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34410         (zero_extendsidi2_dext): Likewise.
34412 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34414         PR sanitizer/80403
34415         * fold-const.c (fold_ternary_loc): Revert
34416         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34417         2017-04-12 change.
34419 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34421         PR rtl-optimization/80343
34422         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34423         new scratch pseudo.
34425 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34427         PR sanitizer/80414
34428         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34429         to ubsan_encode_value.
34431 2017-04-13  Jeff Law  <law@redhat.com>
34433         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34434         appearing in DEBUG_INSNs.
34436 2017-04-13  Martin Liska  <mliska@suse.cz>
34438         PR gcov-profile/80413
34439         * gcov-io.c (gcov_write_string): Copy to buffer just when
34440         allocated size is greater than zero.
34442 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34444         PR debug/80321
34445         * dwarf2out.c (decls_for_scope): Ignore declarations of
34446         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34448 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34450         PR lto/69953
34451         * ipa-visibility.c (non_local_p): Fix typos.
34452         (localize_node): When localizing symbol in same comdat group,
34453         dissolve the group only when we know external symbols are going
34454         to be privatized.
34455         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34457 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34459         PR tree-optimization/79390
34460         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34461         order does not result in usable sequence, retry with reversed operand
34462         order.
34464         PR sanitizer/80403
34465         PR sanitizer/80404
34466         PR sanitizer/80405
34467         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34468         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34469         op0 instead of fold_convert_loc (loc, type, arg0).
34471 2017-04-12  Jeff Law  <law@redhat.com>
34473         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34474         has a delay slot in the generated code.
34476         * config/cris/cris.md (cris_preferred_reload_class): Return
34477         GENNONACR_REGS rather than GENERAL_REGS.
34479 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34481         PR c/80163
34482         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34483         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34484         signedness of the result type.
34486 2017-04-12  Richard Biener  <rguenther@suse.de>
34487             Jeff Law  <law@redhat.com>
34489         PR tree-optimization/80359
34490         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34491         trim stores to TARGET_MEM_REFs.
34493 2017-04-12  Richard Biener  <rguenther@suse.de>
34495         PR tree-optimization/79390
34496         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34497         threading case even more.
34499 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34501         PR target/80382
34502         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34503         for quad_address_p for TImode, instead of just not indexed_address.
34505 2017-04-12  Richard Biener  <rguenther@suse.de>
34506             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34508         PR middle-end/79671
34509         * alias.c (component_uses_parent_alias_set_from): Handle
34510         TYPE_TYPELESS_STORAGE.
34511         (get_alias_set): Likewise.
34512         * tree-core.h (tree_type_common): Add typeless_storage flag.
34513         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34514         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34515         for types containing members with TYPE_TYPELESS_STORAGE.
34516         (place_field): Likewise.
34517         (layout_type): Likewise for ARRAY_TYPE.
34518         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34519         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34520         TYPE_TYPELESS_STORAGE.
34521         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34523 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34525         PR sanitizer/80349
34526         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34527         first argument to type.
34529 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34531         PR target/80376
34532         PR target/80315
34533         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34534         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34535         (rs6000_expand_binop_builtin): Likewise.
34536         (rs6000_expand_ternop_builtin): Likewise; also add missing
34537         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34538         vshasigma built-ins.
34539         * doc/extend.texi: Document that vec_xxpermdi's third argument
34540         must be a constant.
34542 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34544         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34545         Use shift_const cost parameter when calculating gain of STV shifts.
34547 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34549         PR rtl-optimization/70478
34550         * lra-constraints.c (process_alt_operands): Check memory for
34551         disfavoring memory insn operand.
34553 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34555         PR middle-end/80100
34556         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34557         left shift in unsigned HOST_WIDE_INT type.
34559         PR rtl-optimization/80385
34560         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34561         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34563         PR libgomp/80394
34564         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34565         if they have any depend clauses.
34567 2017-04-11  Martin Liska  <mliska@suse.cz>
34569         PR ipa/80212
34570         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34571         * ipa-split.c (split_function): Create a local comdat symbol
34572         if caller is in a comdat group.
34574 2017-04-11  Martin Liska  <mliska@suse.cz>
34576         PR ipa/80212
34577         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34578         flags.
34580 2017-04-11  Martin Sebor  <msebor@redhat.com>
34582         PR middle-end/80364
34583         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34584         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34585         for INTEGER_TYPE.
34586         (directive::set_width, directive::set_precision, format_character):
34587         Adjust.
34588         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34589         INTEGER_TYPE.
34591 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34593         PR target/80389
34594         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34595         conflict, set target->arch_name instead of target->cpu_name.
34597 2017-04-11  Richard Biener  <rguenther@suse.de>
34599         PR tree-optimization/80374
34600         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34601         build_zero_cst, remove fold_convertible_p check again.
34603 2017-04-11  Martin Liska  <mliska@suse.cz>
34605         PR sanitizer/70878
34606         * ubsan.c (instrument_object_size): Do not instrument register
34607         variables.
34609 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34611         PR target/80381
34612         * config/i386/i386-builtin-types.def
34613         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34614         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34615         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34616         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34617         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34618         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34619         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34620         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34621         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34622         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34623         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34624         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34625         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34626         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34627         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34628         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34629         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34630         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34631         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34632         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34633         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34634         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34635         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34636         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34637         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34638         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34639         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34640         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34641         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34642         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34643         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34644         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34645         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34646         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34647         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34648         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34649         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34650         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34651         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34652         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34653         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34654         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34655         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34656         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34657         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34658         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34659         aliases.
34660         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34661         flag to second_arg_count, handle 4 argument function type _COUNT
34662         aliases, handle second_arg_count on second argument rather than last.
34664 2017-04-10  Jeff Law  <law@redhat.com>
34666         PR tree-optimization/80374
34667         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34668         record anything if we can not convert integer_zero_node to the
34669         desired type.
34671 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34673         PR target/80108
34674         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34675         Enhance special handling given to the TARGET_P9_MINMAX option in
34676         relation to certain other options.
34678 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34680         PR tree-optimization/80153
34681         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34682         remove POINTER_PLUS_EXPR's base part directly, rather than through
34683         aff_tree.
34685 2017-04-10  Richard Biener  <rguenther@suse.de>
34686             Bin Cheng  <bin.cheng@arm.com>
34688         PR tree-optimization/80153
34689         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34690         the first element of pointer type aff_tree.  Build result expr in
34691         aff_tree's type.
34692         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34693         fold_convert calls.
34694         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34695         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34697 2017-04-10  Richard Biener  <rguenther@suse.de>
34699         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34700         asm inputs.
34702 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34704         PR rtl-optimization/70478
34705         * lra-constraints.c (curr_small_class_check): New.
34706         (update_and_check_small_class_inputs): New.
34707         (process_alt_operands): Update curr_small_class_check.  Disfavor
34708         alternative insn memory operands.  Check available regs for small
34709         class operands.
34711 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34713         PR target/80057
34714         * config/mips/mips.opt (-mvirt): Update description.
34715         * doc/invoke.texi (-mvirt): Likewise.
34717 2017-04-10  Richard Biener  <rguenther@suse.de>
34719         PR middle-end/80362
34720         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34721         looking for NEGATE_EXPR in -A / -B to A / B folding.
34723 2017-04-10  Martin Liska  <mliska@suse.cz>
34725         PR gcov-profile/80224
34726         * gcov.c (print_usage): Fix usage string.
34727         (get_gcov_intermediate_filename): Remove.
34728         (output_gcov_file): Use both for normal and intermediate format.
34729         (generate_results): Do not initialize special file for
34730         intermediate format.
34732 2017-04-10  Richard Biener  <rguenther@suse.de>
34734         PR tree-optimization/80304
34735         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34736         for safelen.
34738 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34740         PR target/79905
34741         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34742         (rs6000_init_builtins): Use it.
34744 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34746         * config/arm/arm.md (<mrc>): Add mode to SET source.
34747         (<mrrc>): Likewise.
34749 2017-04-10  Richard Biener  <rguenther@suse.de>
34751         PR middle-end/80344
34752         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34754 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34756         PR target/80324
34757         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34758         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34759         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34760         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34761         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34762         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34763         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34764         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34765         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34766         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34767         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34768         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34769         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34770         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34771         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34772         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34773         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34774         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34775         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34776         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34777         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34778         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34779         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34781 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34783         PR rtl-optimization/70478
34784         * lra-constraints.c: Reverse the last patch.
34786 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34788         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34789         Add comment for WCHAR_T.
34791 2017-04-08  Martin Liska  <mliska@suse.cz>
34793         Revert:
34794         2017-04-07  Martin Liska  <mliska@suse.cz>
34796         PR ipa/80212
34797         * ipa-split.c (split_function): Add function part to a same comdat
34798         group.
34800 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34802         PR target/80358
34803         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34805 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34807         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34808         (vec_pairing): ...to this.
34809         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34810         (rs6000_sched_init): Adjust for name change.
34811         (struct rs6000_sched_context): Likewise.
34812         (rs6000_init_sched_context): Likewise.
34813         (rs6000_set_sched_context): Likewise.
34815 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34817         PR target/80322
34818         PR target/80323
34819         PR target/80325
34820         PR target/80326
34821         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34822         intrinsics.
34823         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34824         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34825         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34827 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34829         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34831 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34833         PR rtl-optimization/70703
34834         * ira-color.c (update_conflict_hard_regno_costs): Use
34835         int64_t instead of HOST_WIDE_INT.
34837 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34839         PR rtl-optimization/70478
34840         * lra-constraints.c (process_alt_operands): Disfavor alternative
34841         insn memory operands.
34843 2017-04-07  Jeff Law  <law@redhat.com>
34845         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34846         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34848 2017-04-07  Martin Liska  <mliska@suse.cz>
34850         PR target/79889
34851         * config/aarch64/aarch64.c (aarch64_process_target_attr):
34852         Show error message instead of an ICE.
34854 2017-04-07  Martin Liska  <mliska@suse.cz>
34856         PR ipa/80212
34857         * ipa-split.c (split_function): Add function part to a same comdat
34858         group.
34860 2017-04-07  Richard Biener  <rguenther@suse.de>
34862         PR middle-end/80341
34863         * tree.c (get_unwidened): Also handle ! for_type case for
34864         INTEGER_CSTs.
34865         * convert.c (do_narrow): Split out from ...
34866         (convert_to_integer_1): ... here.  Do not pass final truncation
34867         type to get_unwidened for TRUNC_DIV_EXPR.
34869 2017-04-07  Richard Biener  <rguenther@suse.de>
34871         * tree-affine.c (wide_int_ext_for_comb): Take type rather
34872         than aff_tree.
34873         (aff_combination_const): Adjust.
34874         (aff_combination_scale): Likewise.
34875         (aff_combination_add_elt): Likewise.
34876         (aff_combination_add_cst): Likewise.
34877         (aff_combination_convert): Likewise.
34878         (add_elt_to_tree): Likewise.  Remove unused argument.
34879         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
34881 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34883         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
34884         definition.
34885         * config/arm/arm.c (arm_default_short_enums): Use
34886         ARM_DEFAULT_SHORT_ENUMS.
34887         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
34889 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
34891         PR debug/80234
34892         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
34893         members with redundant out-of-class redeclaration.
34895 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34897         PR target/80286
34898         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
34899         * config/i386/i386.md (*zero_extendsidi2):
34900         Add (?*x,*x) and (?*v,*v) alternatives.
34902 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34904         PR target/79733
34905         * config/i386/i386.c (ix86_expand_builtin)
34906         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
34907         mode from insn data. Convert operands to insn operand mode.
34908         Copy operands that don't satisfy insn predicate to a register.
34910 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
34912         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
34913         Update comments.
34915 2017-04-06  Richard Biener  <rguenther@suse.de>
34917         PR tree-optimization/80334
34918         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
34919         preserve alignment of accesses.
34921 2017-04-06  Richard Biener  <rguenther@suse.de>
34923         PR tree-optimization/80262
34924         * tree-sra.c (build_ref_for_offset): Preserve address-space
34925         information.
34926         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
34927         Drop useless address-space information on MEM_REF offsets.
34929 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
34931         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
34933 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
34935         PR rtl-optimization/70703
34936         * ira-color.c (update_conflict_hard_regno_costs): Use
34937         HOST_WIDE_INT instead of long.
34939 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
34941         PR target/80298
34942         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
34943         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
34944         is not defined.
34945         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
34946         for x86_64 target.  Handle -m3dnowa option.
34948 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
34950         PR rtl-optimization/70703
34951         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
34952         (update_conflict_hard_regno_costs): Use long instead of unsigned
34953         arithmetic for cost calculation.
34955 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
34956             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34958         PR sanitizer/80308
34959         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
34960         for big endian.
34962 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
34964         PR target/78002
34965         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
34966         ptr_mode with Pmode throughout.
34967         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
34968         into probe_stack_range and use DImode.
34970 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34972         PR target/79890
34973         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
34974         call_eh_return is true.
34976 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34978         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
34979         Initialize last_match_fntype_index.
34981 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
34983         PR target/80310
34984         * tree-nvr.c: Include internal-fn.h.
34985         (pass_return_slot::execute): Ignore internal calls without
34986         direct optab.
34988 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
34989             Richard Biener  <rguenther@suse.de>
34991         PR c++/80297
34992         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
34993         captures used multiple times, except for the last use.
34994         * generic-match-head.c: Include gimplify.h.
34996 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
34998         PR tree-optimization/79390
34999         * target.h (struct noce_if_info): Declare.
35000         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35001         * target.def (noce_conversion_profitable_p): New target hook.
35002         * ifcvt.h (struct noce_if_info): New type, moved from ...
35003         * ifcvt.c (struct noce_if_info): ... here.
35004         (noce_conversion_profitable_p): Renamed to ...
35005         (default_noce_conversion_profitable_p): ... this.  No longer
35006         static nor inline.
35007         (noce_try_store_flag_constants, noce_try_addcc,
35008         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35009         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35010         instead of noce_conversion_profitable_p.
35011         * config/i386/i386.c: Include ifcvt.h.
35012         (ix86_option_override_internal): Don't override
35013         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35014         (ix86_noce_conversion_profitable_p): New function.
35015         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35016         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35017         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35018         * doc/tm.texi: Regenerated.
35020 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35022         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35023         correction.
35025 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35027         PR target/80307
35028         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35029         instructions for small multiply cores.
35031 2017-04-04  Jeff Law  <law@redhat.com>
35033         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35034         added member.
35035         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35036         that are not set by the loop over the elements.
35038 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35040         PR target/80286
35041         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35042         int mode, convert_modes it to mode as unsigned, otherwise use
35043         lowpart_subreg to mode rather than SImode.
35044         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35045         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35046         Use DImode instead of SImode for the shift count operand.
35047         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35048         Likewise.
35050 2017-04-04  Richard Biener  <rguenther@suse.de>
35052         PR middle-end/80281
35053         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35054         arithmetic done for the negate or the plus.  Simplify.
35055         (A - (-B) -> A + B): Likewise.
35056         * fold-const.c (split_tree): Make sure to not negate pointers.
35058 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35060         PR rtl-optimization/60818
35061         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35062         a compare of comparisons with the thing compared if this results
35063         in a different machine mode.
35065 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35067         * alias.c (base_alias_check): Fix typo in comment.
35068         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35069         * cgraphunit.c (symbol_table::compile): Likewise.
35070         * collect2.c (maybe_run_lto_and_relink): Likewise.
35071         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35072         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35073         * config/avr/avr.c (avr_map_op_t): Likewise.
35074         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35075         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35076         * config/epiphany/epiphany.md (movcc): Likewise.
35077         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35078         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35079         Likewise.
35080         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35081         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35082         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35083         * config/sh/sh.c (sh_rtx_costs): Likewise.
35084         * fold-const.c (fold_truth_andor): Likewise.
35085         * genautomata.c (collapse_flag): Likewise.
35086         * gengtype.h (struct type::u::s): Likewise.
35087         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35088         * input.c (FORMAT_AMOUNT): Likewise.
35089         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35090         (known_aggs_to_agg_replacement_list): Likewise.
35091         * ipa-inline-analysis.c: Likewise.
35092         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35093         * ipa-polymorphic-call.c
35094         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35095         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35096         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35097         Likewise.
35098         * modulo-sched.c (apply_reg_moves): Likewise.
35099         * omp-expand.c (build_omp_regions_1): Likewise.
35100         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35101         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35102         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35103         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35104         * value-prof.c: Likewise.
35105         * var-tracking.c (val_reset): Likewise.
35107 2017-04-03  Richard Biener  <rguenther@suse.de>
35109         PR tree-optimization/80275
35110         * fold-const.c (split_address_to_core_and_offset): Handle
35111         POINTER_PLUS_EXPR.
35113 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35115         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35116         descriptors is at least equal to that of functions.
35118 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35120         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35122 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35124         PR target/80250
35125         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35126         (mov<IMOD4:mode>): New expander.
35127         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35129 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35131         PR rtl-optimization/79405
35132         * fwprop.c (propagations_left): New variable.
35133         (forward_propagate_into): Decrement it.
35134         (fwprop_init): Initialize it.
35135         (fw_prop): If the variable has reached zero, stop propagating.
35136         (fwprop_addr): Ditto.
35138 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35140         PR debug/79255
35141         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35142         a FUNCTION_DECL, pass it as decl instead of origin to
35143         process_scope_var.
35145 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35147         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35148         string.
35150 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35152         PR target/80107
35153         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35154         TARGET_VSX_SMALL_INTEGER.
35156 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35158         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35159         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35161 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35163         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35164         extraction from odd-numbered MSA register.
35166 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35168         PR middle-end/80173
35169         * expmed.c (store_bit_field_1): Don't attempt to create
35170         a word subreg out of hard registers wider than word if they
35171         have HARD_REGNO_NREGS of 1 for their mode.
35173         PR middle-end/80163
35174         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35175         conversions to integer types wider than word and pointer.
35177         PR debug/80025
35178         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35179         (rtx_equal_for_cselib_p): Pass 0 to it.
35180         * cselib.c (cselib_hasher::equal): Likewise.
35181         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35182         is 128, don't look up VALUE locs and punt.  Increment
35183         depth in recursive calls when walking VALUE locs.
35185 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35187         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35188         (make_gcov_file_name): Use the canonical path name for generating
35189         the MD5 value.
35190         (read_line): Fix handling of files with ascii null bytes.
35192 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35194         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35195         to initialise a vector register instead
35196         of using a const_int.
35198 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35200         PR translation/80189
35201         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35202         diagnostic messages.
35204 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35206         PR target/80246
35207         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35208         (dfp_diex_<mode>): Update mode of operand 1.
35209         * doc/extend.texi (dxex, dxexq): Document change to return type.
35210         (diex, diexq): Document change to argument type.
35212 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35214         PR ipa/77333
35215         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35216         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35217         it reflects the signature changes performed at the callee side.
35218         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35219         to cgraph_build_function_type_skip_args.
35220         (build_function_decl_skip_args): Adjust call to the above function.
35222 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35224         PR target/80206
35225         * config/i386/sse.md
35226         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35227         register as dest whenever it is a MEM not rtx_equal_p to the
35228         corresponding dup operand, and when forcing into reg move the
35229         reg into the memory afterwards.
35230         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35231         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35232         for the force_reg mode.
35233         (avx512vl_vextractf128<mode>): Use register as dest either
35234         always when a MEM, or when it is a MEM not rtx_equal_p to the
35235         corresponding dup operand, or even not when it is a CONST_VECTOR
35236         depending on the mode and lo vs. hi.
35237         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35238         parens.
35239         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35240         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35241         Likewise.  Require that operands[2] is even.
35242         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35243         Remove extraneous parens.  Require that operands[2] is a multiple
35244         of 4.
35245         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35246         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35247         disallow memory then.
35249 2017-03-30  Richard Biener  <rguenther@suse.de>
35251         PR tree-optimization/77498
35252         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35253         to non-constants over backedges.
35255 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35257         PR rtl-optimization/80233
35258         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35259         as last_combined_insn.  Do not test for BARRIER_P separately.
35261 2017-03-29  Andreas Schwab  <schwab@suse.de>
35263         PR ada/80146
35264         * calls.c (prepare_call_address): Convert funexp to Pmode before
35265         copying to temp reg.
35267 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35269         PR tree-optimization/80158
35270         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35271         Handle possible future case of more than one alternate
35272         interpretation.
35273         (replace_rhs_if_not_dup): Likewise.
35274         (replace_one_candidate): Likewise.
35276 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35278         PR rtl-optimization/80193
35279         * ira.c (ira): Do not check allocation for LRA.
35281 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35283         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35284         (nvptx_output_simt_exit): Declare.
35285         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35286         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35287         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35288         (nvptx_declare_function_name): ...here.  Emit declaration of local
35289         memory space buffer for omp_simt_enter insn.
35290         (nvptx_output_unisimt_switch): New.
35291         (nvptx_output_softstack_switch): New.
35292         (nvptx_output_simt_enter): New.
35293         (nvptx_output_simt_exit): New.
35294         * config/nvptx/nvptx.h (struct machine_function): New fields
35295         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35296         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35297         (UNSPECV_SIMT_EXIT): Ditto.
35298         (omp_simt_enter_insn): New insn.
35299         (omp_simt_enter): New expansion.
35300         (omp_simt_exit): New insn.
35301         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35303         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35304         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35305         (expand_GOMP_SIMT_EXIT): New.
35306         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35307         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35308         (GOMP_SIMT_EXIT): Ditto.
35309         * target-insns.def (omp_simt_enter): New insn.
35310         (omp_simt_exit): Ditto.
35311         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35312         simt_dlist.
35313         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35314         (lower_rec_input_clauses): Likewise.
35315         (lower_lastprivate_clauses): Handle SIMT privatization.
35317         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35318         (ompdevlow_adjust_simt_enter): New.
35319         (find_simtpriv_var_op): New.
35320         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35321         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35323         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35324         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35325         (copy_decl_for_dup_finish): Ditto.
35327         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35329 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35331         PR target/53383
35332         * config/i386/i386.c (ix86_option_override_internal): Always
35333         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35335 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35337         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35339 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35341         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35342         mark new edge's irreducible flag accordign to it.
35343         (vect_do_peeling): Check loop preheader edge's irreducible flag
35344         and pass it to function slpeel_add_loop_guard.
35346 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35348         PR tree-optimization/80218
35349         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35350         Update block frequencies and counts.
35352 2017-03-28  Richard Biener  <rguenther@suse.de>
35354         PR tree-optimization/78644
35355         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35356         of a simplification result we may not use it at all.
35358 2017-03-28  Richard Biener  <rguenther@suse.de>
35360         PR ipa/80205
35361         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35362         without arguments, generate default definition of a SSA name.
35364 2017-03-28  Richard Biener  <rguenther@suse.de>
35366         PR middle-end/80222
35367         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35368         TYPE_REF_CAN_ALIAS_ALL references.
35369         * fold-const.c (fold_indirect_ref_1): Likewise.
35371 2017-03-28  Martin Liska  <mliska@suse.cz>
35373         PR ipa/80104
35374         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35375         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35377 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35378             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35380         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35381         (EXTRA_SPECS): Define.
35382         (SUBTARGET_EXTRA_SPECS): Likewise.
35383         (SUBTARGET_CPP_SPEC): Likewise.
35384         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35385         SUBTARGET_EXTRA_SPECS.
35386         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35388 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35390         * config/arc/simdext.md (vst64_insn): Update pattern.
35391         (vld32wh_insn): Likewise.
35392         (vld32wl_insn): Likewise.
35393         (vld64_insn): Likewise.
35394         (vld32_insn): Likewise.
35396 2017-03-28  Marek Polacek  <polacek@redhat.com>
35398         PR sanitizer/80067
35399         * fold-const.c (fold_comparison): Use protected_set_expr_location
35400         instead of SET_EXPR_LOCATION.
35402 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35404         * tree.c (add_expr): Avoid name lookup warning.
35406 2017-03-27  Jeff Law  <law@redhat.com>
35408         PR tree-optimization/80216
35409         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35410         function name.  Limit recursion depth.
35411         (record_temporary_equivalences): Corresponding changes.
35413 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35415         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35416         covered first.
35418 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35420         PR target/80102
35421         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35422         notes.
35423         * cfgcleanup.c (reg_note_cfa_p): New array.
35424         (insns_have_identical_cfa_notes): New function.
35425         (old_insns_match_p): Don't cross-jump in between /f
35426         and non-/f instructions.  If both i1 and i2 are frame related,
35427         verify all CFA notes, their order and content.
35429 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35431         PR target/78543
35432         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35433         HImode and SImode with zero extend to DImode to one insn.
35434         (bswap<mode>2_extenddi): Likewise.
35435         (bswapsi2_extenddi): Likewise.
35436         (bswaphi2_extendsi): Likewise.
35437         (bswaphi2): Combine bswap HImode and SImode into one insn.
35438         Separate memory insns from swapping register.
35439         (bswapsi2): Likewise.
35440         (bswap<mode>2): Likewise.
35441         (bswaphi2_internal): Delete, no longer used.
35442         (bswapsi2_internal): Likewise.
35443         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35444         store, and gpr<-gpr swap insns.
35445         (bswap<mode>2_store): Likewise.
35446         (bswaphi2_reg): Register only splitter, combine with the splitter.
35447         (bswaphi2 splitter): Likewise.
35448         (bswapsi2_reg): Likewise.
35449         (bswapsi2 splitter): Likewise.
35450         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35451         the insns into load, store, and register/register insns.
35452         (bswapdi2_ldbrx): Likewise.
35453         (bswapdi2_load): Likewise.
35454         (bswapdi2_store): Likewise.
35455         (bswapdi2_reg): Likewise.
35457 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35459         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35460         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35462 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35464         PR target/80103
35465         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35466         add comments.
35467         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35468         special handling for target option conflicts between dform
35469         options (-mpower9-dform, -mpower9-dform-vector,
35470         -mpower9-dform-scalar) and -mno-direct-move.
35472 2017-03-27  Richard Biener  <rguenther@suse.de>
35474         PR tree-optimization/80181
35475         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35477 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35479         * config/arc/predicates.md (move_double_src_operand): Replace the
35480         call to move_double_src_operand with a call to address_operand.
35482 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35484         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35485         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35486         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35488 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35490         * config/arc/predicates.md (long_immediate_loadstore_operand):
35491         Consider scaled addresses cases.
35493 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35495         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35496         restored when in interrupt.
35497         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35498         doesn't have delay slot.
35500 2017-03-27  Richard Biener  <rguenther@suse.de>
35502         PR ipa/79776
35503         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35504         inlined thunk clones.
35506 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35508         PR sanitizer/80168
35509         * asan.c (instrument_derefs): Copy over last operand from
35510         original COMPONENT_REF to the new COMPONENT_REF with
35511         DECL_BIT_FIELD_REPRESENTATIVE.
35512         * ubsan.c (instrument_object_size): Likewise.
35514 2017-03-27  Richard Biener  <rguenther@suse.de>
35516         PR tree-optimization/80170
35517         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35518         sure DR/SCEV didnt fold in constants we do not see when looking
35519         at the reference base alignment.
35521 2017-03-27  Richard Biener  <rguenther@suse.de>
35523         PR middle-end/80171
35524         * gimple-fold.c (fold_ctor_reference): Properly guard against
35525         NULL return value from canonicalize_constructor_val.
35527 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35529         PR target/80180
35530         * config/i386/i386.c (ix86_expand_builtin)
35531         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35532         flags reg setting and flags reg using instructions.
35533         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35534         clobbering instructions to zero extend op2.
35536 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35538         * doc/install.texi (Configuration) <--with-aix-soname>:
35539         Update link to AIX ld.
35541 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35543         PR rtl-optimization/80160
35544         PR rtl-optimization/80159
35545         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35546         reg_alternate_class into account.
35548 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35550         PR target/80148
35551         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35552         to consider in curr_insn_transform.
35554 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35556         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35557         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35558         and emit_mode_inner.
35560 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35562         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35563         argument to the overloaded builtin variants.  Use the new flag to
35564         deprecate certain builtin variants.
35565         * config/s390/s390-builtin-types.def: Add new builtin types.
35566         * config/s390/s390-builtins.h: Support new flags field for
35567         overloaded builtins.
35568         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35569         (s390_macro_to_expand): Enable vector float data type.
35570         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35571         builtins by incrementing the __VEC__ version number.
35572         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35573         vec_xst.
35574         (s390_resolve_overloaded_builtin): Emit error messages depending
35575         on the builtin flags.
35576         * config/s390/s390.c (s390_expand_builtin): Support additional
35577         flags argument.  Change error message to match the messages
35578         emitted in s390-c.c.
35579         * config/s390/s390.md: New UNSPEC_* constants.
35580         (op_type): Add new instruction types.
35581         * config/s390/vecintrin.h: Add new builtins and test data class
35582         constants.
35583         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35584         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35585         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35586         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35587         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35588         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35590         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35591         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35592         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35593         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35595         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35596         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35597         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35598         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35600         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35601         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35602         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35603         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35604         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35605         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35606         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35608         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35609         ("vec_scatter_element<V_HW_4:mode>_DI")
35610         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35611         ("vec_fpint<mode>", "vflls")
35612         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35613         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35614         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35615         ("*vec_cmphe<mode>_cc"): ... these.
35617         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35618         mode constant instead of magic value.
35620 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35622         * config/s390/s390.c (s390_expand_vec_compare): Support other
35623         vector floating point modes than just V2DF.
35624         (s390_expand_vcond): Likewise.
35625         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35626         (s390_cannot_change_mode_class): Prevent mode changes between TF
35627         and V1TF in vector registers.
35628         * config/s390/s390.md (DF, SF): New mode attributes.
35629         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35630         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35631         SFmode support for VRs.
35632         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35633         vector fp modes.
35634         (VFT, VF_HW): New mode iterators.
35635         (vw, sdx): New mode attributes.
35636         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35637         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35638         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35639         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35640         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35641         also the new vector floating point modes.  Renaming to ...
35643         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35644         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35645         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35646         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35647         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35648         ("vec_unordered<mode>"): ... these.
35650         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35651         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35652         ("*vec_extendv2df"): New insn definitions.
35654 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35656         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35657         ("mulditi3_2", "*muldi3_sign"): New patterns.
35658         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35659         rename the pattern definition.
35661 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35663         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35664         expander.
35665         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35667 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35669         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35670         instruction if possible.
35671         * config/s390/vector.md (vec_halfnumelts): New mode
35672         attribute.
35673         ("*vec_vllezlf<mode>"): New pattern.
35675 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35677         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35678         ("popcountv4si2", "popcountv2di2"): Rename to ...
35679         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35680         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35681         condition.
35682         ("popcount<mode>2_vxe"): New pattern.
35684 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35686         * common/config/s390/s390-common.c (processor_flags_table): Add
35687         arch12.
35688         * config.gcc: Add arch12.
35689         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35690         Default to arch12 for unknown CPU model numbers.
35691         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35692         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35693         PROCESSOR_max sanity check.
35694         * config/s390/s390-opts.h (enum processor_type): Add
35695         PROCESSOR_ARCH12.
35696         * config/s390/s390.c (processor_table): Add arch12.
35697         (s390_expand_builtin): Add check for B_VXE flag.
35698         (s390_issue_rate): Add PROCESSOR_ARCH12.
35699         (s390_get_sched_attrmask): Likewise.
35700         (s390_get_unit_mask): Likewise.
35701         (s390_sched_score): Enable z13 scheduling for arch12.
35702         (s390_sched_reorder): Likewise.
35703         (s390_sched_variable_issue): Likewise.
35704         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35705         PF_VXE.
35706         (s390_tune_attr): Use z13 scheduling also for arch12.
35707         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35708         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35709         (TARGET_VXE_P): New macros.
35710         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35711         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35712         * config/s390/s390.opt: Add arch12 as processor_type.
35714 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35716         * config/s390/s390.md
35717         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35718         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35719         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35721         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35722         Rename expanders to ...
35724         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35725         ("fixuns_truncdddi2_emu"): ... these.
35727         ("fixuns_trunc<mode>si2_emu"): New expander.
35729         ("*fixuns_truncdfdi2_z13"): Rename to ...
35730         ("*fixuns_truncdfdi2_vx"): ... this.
35732 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35734         * config/s390/2964.md: Remove the single element vector compare
35735         instructions which are no longer used.
35736         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35737         vector CCmodes.
35738         (s390_canonicalize_comparison): Remove handling of DFmode
35739         compares.
35740         (s390_expand_vec_compare_scalar): Remove function.
35741         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35742         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35743         pattern.
35744         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35746 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35748         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35749         FP zero.
35750         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35751         will anyway by matched by mov<mode>_64dfp.
35753 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35755         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35756         vlef/vstef.  Add missing operand to vleif.
35758 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35760         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35761         pair for all vector types with 64 bit elements.
35762         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35763         * config/s390/vector.md (V_HW_64): ... here.
35764         (V_128_NOSINGLE): New mode iterator.
35765         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35766         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35767         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35768         ("*vec_load_pairv2di"): Change to ...
35769         ("*vec_load_pair<mode>"): ... this one.
35771 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35773         * config/s390/constraints.md: Add comments.
35774         (jKK): Reject element sizes > 8 bytes.
35775         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35776         s_operands.
35777         * config/s390/s390.md: Add the s_operand checks formerly in
35778         s390_split_ok_p to various splitters where they are still
35779         required.
35780         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35781         for 128 bit vectors.  Plus two splitters.
35783 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35785         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35786         the file.
35788 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35790         PR target/79893
35791         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35792         error if the boundary argument is not constant.
35794 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35796         PR rtl-optimization/80112
35797         * loop-doloop.c (doloop_condition_get): Don't check condition
35798         if cmp isn't SET with IF_THEN_ELSE src.
35800 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35802         PR tree-optimization/80158
35803         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35804         replacing a candidate statement, also replace it for the
35805         candidate's alternate interpretation.
35806         (replace_rhs_if_not_dup): Likewise.
35807         (replace_one_candidate): Likewise.
35809 2017-03-24  Richard Biener  <rguenther@suse.de>
35811         PR tree-optimization/80167
35812         * graphite-isl-ast-to-gimple.c
35813         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35814         properly.
35815         (translate_isl_ast_to_gimple::get_rename): Likewise.
35817 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35819         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35820         handling of certain combinations of target options, including the
35821         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35822         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35824 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35826         PR target/71436
35827         * config/arm/arm.md (*load_multiple): Add reload_completed to
35828         matching condition.
35830 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35831             Richard Biener  <rguenth@suse.de>
35833         PR tree-optimization/79908
35834         PR tree-optimization/80136
35835         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35836         been cast away, gimplify_and_add suffices.
35838 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35840         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35842 2017-03-23  Richard Biener  <rguenther@suse.de>
35844         PR tree-optimization/80032
35845         * gimplify.c (gimple_push_cleanup): Forced unconditional
35846         cleanups still have to go to the conditional_cleanups
35847         sequence.
35849 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
35851         PR tree-optimization/80072
35852         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35853         to unsigned int.
35854         (next_operand_entry_id): Change type to unsigned int.
35855         (sort_by_operand_rank): Make sure to return the right return value
35856         even if unsigned fields are bigger than INT_MAX.
35857         (struct oecount): Change cnt and id type to unsigned int.
35858         (oecount_hasher::equal): Formatting fix.
35859         (oecount_cmp): Make sure to return the right return value
35860         even if unsigned fields are bigger than INT_MAX.
35861         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
35863         PR c++/80129
35864         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
35865         TREE_READONLY on result if writing it more than once.
35867         PR sanitizer/80110
35868         * doc/invoke.texi (-fsanitize=thread): Document that with
35869         -fnon-call-exceptions atomics are not able to throw
35870         exceptions.
35872         PR sanitizer/80110
35873         * tsan.c: Include tree-eh.h.
35874         (instrument_builtin_call): Call maybe_clean_eh_stmt or
35875         maybe_clean_or_replace_eh_stmt where needed.
35876         (instrument_memory_accesses): Add cfg_changed argument.
35877         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
35878         if it returned true.
35879         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
35881         PR rtl-optimization/63191
35882         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
35883         wrapper function, moved the whole old content into ...
35884         (ix86_delegitimize_address_1): ... this.  New inline function.
35885         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
35886         true as last argument instead of ix86_delegitimize_address.
35888 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35890         * config/aarch64/aarch64.c (generic_branch_cost): Copy
35891         cortexa57_branch_cost.
35893 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35895         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
35897 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35899         PR target/80123
35900         * doc/md.texi (Constraints): Document wA constraint.
35901         * config/rs6000/constraints.md (wA): New.
35902         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
35903         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
35904         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
35905         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
35907 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
35909         PR c++/80029
35910         * gimplify.c (is_oacc_declared): New function.
35911         (oacc_default_clause): Use it to set default flags for acc declared
35912         variables inside parallel regions.
35913         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
35914         declared variables.
35915         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
35916         declare attribute to any decl as necessary.
35918 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35920         PR target/80082
35921         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
35922         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
35923         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
35924         (arm_arch_lpae): This.
35925         * config/arm/arm.c (arm_arch7ve): Rename into ...
35926         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
35927         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
35928         arm_arch_lpae.
35930 2017-03-22  Martin Liska  <mliska@suse.cz>
35932         PR target/79906
35933         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
35934         error message instead of an ICE.
35936 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35938         * doc/extend.texi (6.11 Additional Floating Types): Revise.
35940 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35942         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
35943         comments.
35944         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35945         comments.
35947 2017-03-21  Martin Sebor  <msebor@redhat.com>
35949         * doc/extend.texi: Use "cannot" instead of "can't."
35950         * doc/hostconfig.texi: Same.
35951         * doc/install.texi: Same.
35952         * doc/invoke.texi: Same.
35953         * doc/loop.texi: Same.
35954         * doc/md.texi: Same.
35955         * doc/objc.texi: Same.
35956         * doc/rtl.texi: Same.
35957         * doc/tm.texi: Same.
35958         * doc/tm.texi.in: Same.
35959         * doc/trouble.texi: Same.
35961 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
35963         PR debug/63238
35964         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
35965         (collect_checksum_attributes): Set it.
35966         (die_checksum_ordered): Use it.
35968 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35970         PR tree-optimization/79908
35971         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
35972         change: For a VA_ARG whose LHS has been cast away, use
35973         force_gimple_operand to construct the side effects.
35975 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
35977         PR translation/80001
35978         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
35979         more amenable to translation.
35980         (oacc_loop_auto_partitions): Likewise.
35982 2017-03-21  Marek Polacek  <polacek@redhat.com>
35983             Martin Sebor  <msebor@redhat.com>
35985         PR tree-optimization/80109
35986         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
35987         on INTEGRAL_TYPE_P.
35989 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
35990             Segher Boessenkool  <segher@kernel.crashing.org>
35992         PR target/80125
35993         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
35994         check reg_used_between_p between insn and one of succ or succ2
35995         depending on if succ is artificial insn not inserted into insn
35996         stream.
35998 2017-03-21  Martin Liska  <mliska@suse.cz>
36000         PR gcov-profile/80081
36001         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36002         * doc/gcc.texi: Include gcov-dump stuff.
36003         * doc/gcov-dump.texi: New file.
36005 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36007         PR rtl-optimization/79150
36008         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36009         conditional jump, if the jump is the last insn of the loop.
36011 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36012             Richard Biener  <rguenth@suse.de>
36014         PR tree-optimization/79908
36015         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36016         been cast away, use force_gimple_operand to construct the side
36017         effects.
36019 2017-03-21  Martin Liska  <mliska@suse.cz>
36021         PR libfortran/79956
36022         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36023         to NULL.
36025 2017-03-21  Brad Spengler <spender@grsecurity.net>
36027         PR plugins/80094
36028         * plugin.c (htab_hash_plugin): New function.
36029         (add_new_plugin): Use it and adjust.
36030         (parse_plugin_arg_opt): Adjust.
36031         (init_one_plugin): Likewise.
36033 2017-03-21  Richard Biener  <rguenther@suse.de>
36035         PR tree-optimization/80032
36036         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36037         if set force the cleanup to happen unconditionally.
36038         (gimplify_target_expr): Push inserted clobbers with force_uncond
36039         to avoid them being removed by control-dependent DCE.
36041 2017-03-21  Richard Biener  <rguenther@suse.de>
36043         PR tree-optimization/80122
36044         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36045         va_arg_pack_len when the inlined call stmt requires pack
36046         expansion itself.
36047         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36049 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36051         PR sanitizer/78158
36052         * tsan.c (instrument_builtin_call): If the memory model argument
36053         is not a constant, assume it is valid.
36055         PR c/67338
36056         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36057         avoid UB.
36059 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36061         PR rtl-optimization/79910
36062         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36063         if its dest is used by an insn before I2 (other than the combined
36064         insns themselves, which are properly handled already).
36066 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36068         Revert:
36069         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36071         * combine.c (record_used_regs): New static function.
36072         (try_combine): Handle situations where there is an additional
36073         instruction between I2 and I3 which needs to have a LOG_LINK
36074         updated.
36076         Revert:
36077         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36079         * combine.c (try_combine): Delete redundant i1 test.  Call
36080         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36082 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36084         PR target/80083
36085         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36086         alternatives 13/14.
36088 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36090         PR tree-optimization/80054
36091         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36092         the optimization if a PHI or any of its arguments is not dominated
36093         by the candidate's basis.  Use gphi* rather than gimple* as
36094         appropriate.
36095         (replace_profitable_candidates): Clean up a gimple* variable that
36096         should be a gphi* variable.
36098 2017-03-20  Martin Sebor  <msebor@redhat.com>
36100         PR c++/52477
36101         * doc/extend.texi (attribute constructor): Document present limitation.
36103 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36105         PR target/79963
36106         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36107         __POWER9_VECTOR__ #ifdef control, change template definition to
36108         use Power9-specific built-in function.
36109         (vec_any_eq): Likewise.
36110         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36111         to control outcomes from this test.
36112         (vector_ae_<mode>p): For VEC_F modes, likewise.
36114 2017-03-20  Ian Lance Taylor  <iant@google.com>
36116         * config/i386/i386.c (ix86_function_regparm): Save an extra
36117         register for -fsplit-stack with DECL_STATIC_CHAIN.
36119 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36121         PR target/79912
36122         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36123         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36125 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36127         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36128         iorw,ow".
36129         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36130         iorw,iorw".
36132 2017-03-20  Marek Polacek  <polacek@redhat.com>
36134         PR sanitizer/80063
36135         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36137 2017-03-20  Richard Biener  <rguenther@suse.de>
36139         PR tree-optimization/80113
36140         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36141         allocate extra SSA name for PHI def.
36142         (add_close_phis_to_outer_loops): Likewise.
36143         (add_close_phis_to_merge_points): Likewise.
36144         (copy_loop_close_phi_args): Likewise.
36145         (copy_cond_phi_nodes): Likewise.
36147 2017-03-20  Martin Liska  <mliska@suse.cz>
36149         PR middle-end/79753
36150         * tree-chkp.c (chkp_build_returned_bound): Do not build
36151         returned bounds for a LHS that's not a BOUNDED_P type.
36153 2017-03-20  Martin Liska  <mliska@suse.cz>
36155         PR target/79769
36156         PR target/79770
36157         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36158         COMPLEX_CST and VECTOR_CST.
36160 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36162         PR target/78857
36163         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36164         target operand.  A new splitter adds the clobber statement in case
36165         the target operand is dead anyway.
36167 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36169         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36170         to age-old versions of binutils and glibc.
36172 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36174         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36176 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36178         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36180 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36182         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36183         requirement for binutils 2.13.
36185 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36187         * combine.c (try_combine): Delete redundant i1 test.  Call
36188         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36190 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36192         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36193         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36194         contents.
36195         <riscv64-*-elf>: Re-arrange section
36196         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36197         <riscv32-*-linux>: Likewise.
36198         <riscv64-*-elf>: Likewise
36199         <riscv64-*-linux>: Likewise.
36201 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36203         PR target/80052
36204         * aarch64.opt(verbose-cost-dump): Fix typo.
36206 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36208         PR target/79951
36209         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36210         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36212 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36214         * reload.c (find_reloads): When reloading a nonoffsettable address,
36215         use RELOAD_OTHER for it and its address reloads.
36217         PR rtl-optimization/79910
36218         * combine.c (record_used_regs): New static function.
36219         (try_combine): Handle situations where there is an additional
36220         instruction between I2 and I3 which needs to have a LOG_LINK
36221         updated.
36223 2017-03-17  Jeff Law  <law@redhat.com>
36225         PR tree-optimization/71437
36226         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36227         conditional in the hash table first.
36228         (vrp_dom_walker::before_dom_children): Extract condition from
36229         ASSERT_EXPR.  Record condition, its inverion and any implied
36230         conditions as well.
36232 2017-03-17  Marek Polacek  <polacek@redhat.com>
36233             Markus Trippelsdorf  <markus@trippelsdorf.de>
36235         PR tree-optimization/80079
36236         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36237         m_stores_head.
36239 2017-03-17  Richard Biener  <rguenther@suse.de>
36241         PR middle-end/80075
36242         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36243         Properly verify the LHS before the RHS possibly claims to be
36244         handled.
36245         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36246         do not throw.
36248 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36250         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36251         (List of -O2 options): Likewise.
36252         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36253         (-fipa-vrp) New.
36255 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36257         * gcov-dump.c (print_usage): Print bug_report_url.
36259 2017-03-17  Richard Biener  <rguenther@suse.de>
36261         PR middle-end/80050
36262         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36263         (parser::peek): Likewise.
36265 2017-03-17  Richard Biener  <rguenther@suse.de>
36267         PR tree-optimization/80048
36268         * sese.c (free_sese_info): Properly release rename_map and
36269         copied_bb_map elements.
36271 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36273         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36274         Add linked-list forward and backlinks.  Insert on
36275         construction, remove on destruction.
36276         (class pass_store_merging): Add m_stores_head field.
36277         (pass_store_merging::terminate_and_process_all_chains):
36278         Iterate over m_stores_head list.
36279         (pass_store_merging::terminate_all_aliasing_chains):
36280         Likewise.
36281         (pass_store_merging::execute): Check for debug stmts first.
36282         Push new chains onto the m_stores_head stack.
36284 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36286         PR target/71294
36287         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36288         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36289         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36291 2017-03-16  Jeff Law  <law@redhat.com>
36293         PR tree-optimization/71437
36294         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36295         member function.  Implementation moved into after_dom_children
36296         member function and into the threader's thread_outgoing_edges
36297         function.
36298         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36299         some code into new thread_outgoing_edges.
36300         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36301         definition.  Simplify marker handling (do it here).   Assume we always
36302         have the available expression and the const/copies tables.
36303         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36304         and tree-vrp.c
36305         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36306         * tree-vrp.c (equiv_stack): No longer file scoped.
36307         (vrp_dom_walker): New class.
36308         (vrp_dom_walker::before_dom_children): New member function.
36309         (vrp_dom_walker::after_dom_children): Likewise.
36310         (identify_jump_threads):  Setup domwalker.  Use it rather than
36311         walking edges in a random order by hand.  Simplify setup/finalization.
36312         (finalize_jump_threads): Remove.
36313         (vrp_finalize): Do not call identify_jump_threads here.
36314         (execute_vrp): Do it here instead and call thread_through_all_blocks
36315         here too.
36317         PR tree-optimization/71437
36318         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36319         callers changed.
36320         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36321         callers changed.
36322         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36323         (dom_opt_dom_walker::thread_across_edge): Remove
36324         handle_dominating_asserts argument.  All callers changed.
36325         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36326         changes.  Remove calls to lhs_of_dominating_assert.  Other
36327         uses of handle_dominating_asserts turn into unconditional code
36328         (simplify_control_stmt_condition_1): Likewise.
36329         (simplify_control_stmt_condition): Likewise.
36330         (thread_through_normal_block, thread_across_edge): Likewise.
36331         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36332         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36333         object if it is not an SSA_NAME.
36334         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36335         before calling into the VRP specific simplifiers.
36336         (identify_jump_threads): Remove handle_dominating_asserts
36337         argument.
36339 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36341         PR fortran/79886
36342         * tree-diagnostic.c (default_tree_printer): No longer static.
36343         * tree-diagnostic.h (default_tree_printer): New prototype.
36345 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36347         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36348         Change ins into fmov.
36350 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36352         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36353         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36354         Use h_con constraint for operand 1.
36355         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36356         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36358 2017-03-15  Jeff Law  <law@redhat.com>
36360         PR tree-optimization/71437
36361         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36362         (record_temporary_equivalences): Use it.
36364         PR tree-optimization/71437
36365         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36366         tree-ssa-scopedtables.
36367         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36368         (record_conditions, record_cond, vuse_eq): Likewise.
36369         (record_edge_info): Adjust to API tweak of record_conditions.
36370         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36371         (record_temporary_equivalences, optimize_stmt): Likewise.
36372         (eliminate_redundant_computations): Likewise.
36373         (record_equivalences_from_stmt): Likewise.
36374         * tree-ssa-scopedtables.c: Include options.h and params.h.
36375         (vuse_eq): New function, moved from tree-ssa-dom.c
36376         (build_and_record_new_cond): Likewise.
36377         (record_conditions): Likewise.  Accept vector of conditions rather
36378         than edge_equivalence structure for first argument.
36379         for the first argument.
36380         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36381         from tree-ssa-dom.c.
36382         (avail_exprs_stack::record_cond): Likewise.
36383         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36384         from tree-ssa-dom.c.
36385         (avail_exprs_stack): Add new member functions lookup_avail_expr
36386         and record_cond.
36387         (record_conditions): Declare.
36389 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36391         PR target/80017
36392         * lra-constraints.c (process_alt_operands): Increase reject for
36393         reloading an input/output operand.
36395 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36397         PR target/79038
36398         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36399         insns to convert from signed/unsigned char/short to IEEE 128-bit
36400         floating point.
36401         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36403 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36405         PR target/80019
36406         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36407         subreg of inner mode for values already in registers.
36409 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36411         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36412         iteration reg is used after the loop.
36414 2017-03-14  Martin Sebor  <msebor@redhat.com>
36416         PR tree-optimization/79800
36417         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36418         precision in negative-positive range.
36419         (format_floating): Call non-const overload with adjusted precision.
36421 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36423         PR target/79947
36424         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36425         -mpowerpc-gfxopt.
36427 2017-03-14  Martin Sebor  <msebor@redhat.com>
36429         PR middle-end/80020
36430         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36431         * builtins.def (aligned_alloc): Use it.
36433         PR c/79936
36434         * Makefile.in (GTFILES): Add calls.c.
36435         * calls.c: Include "gt-calls.h".
36437 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36439         PR rtl-optimization/79728
36440         * regs.h (struct target_regs): New field
36441         x_contains_allocatable_regs_of_mode.
36442         (contains_allocatable_regs_of_mode): New macro.
36443         * reginfo.c (init_reg_sets_1): Initialize it, and change
36444         contains_reg_of_mode so it includes global regs as well.
36445         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36446         rather than contains_regs_of_mode.
36448 2017-03-14  Martin Liska  <mliska@suse.cz>
36450         * doc/invoke.texi: Document options that can't be combined with
36451         -fcheck-pointer-bounds.
36453 2017-03-14  Martin Liska  <mliska@suse.cz>
36455         PR middle-end/79831
36456         * doc/invoke.texi (-Wchkp): Document the option.
36458 2017-03-14  Martin Liska  <mliska@suse.cz>
36460         * Makefile.in: Install gcov-dump.
36462 2017-03-14  Martin Liska  <mliska@suse.cz>
36464         * multiple_target.c (expand_target_clones): Bail out for
36465         an invalid attribute.
36467 2017-03-14  Richard Biener  <rguenther@suse.de>
36469         * alias.c (struct alias_set_entry): Pack properly.
36470         * cfgloop.h (struct loop): Likewise.
36471         * cse.c (struct set): Likewise.
36472         * ipa-utils.c (struct searchc_env): Likewise.
36473         * loop-invariant.c (struct invariant): Likewise.
36474         * lra-remat.c (struct cand): Likewise.
36475         * recog.c (struct change_t): Likewise.
36476         * rtl.h (struct address_info): Likewise.
36477         * symbol-summary.h (function_summary): Likewise.
36478         * tree-loop-distribution.c (struct partition): Likewise.
36479         * tree-object-size.c (struct object_size_info): Likewise.
36480         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36481         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36482         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36483         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36484         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36485         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36486         (struct _stmt_vec_info): Likewise.
36488 2017-03-14  Martin Liska  <mliska@suse.cz>
36490         PR target/79892
36491         * multiple_target.c (create_dispatcher_calls): Check that
36492         a target can create a function dispatcher.
36494 2017-03-14  Martin Liska  <mliska@suse.cz>
36496         PR lto/66295
36497         * multiple_target.c (expand_target_clones): Drop local.local
36498         flag for default implementation.
36500 2017-03-14  Richard Biener  <rguenther@suse.de>
36502         PR tree-optimization/80030
36503         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36505 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36507         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36508         gcc_fallthrough() instead of __attribute__((fallthrough));
36510 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36512         * doc/gcc.texi: Remove "up" link to (DIR).
36513         * doc/gccint.texi: Ditto.
36515 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36517         * doc/install.texi (Specific) <avr>: Remove reference to
36518         binutils 2.13.
36520 2017-03-13  Jeff Law  <law@redhat.com>
36522         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36523         attribute rather than comments.
36525         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36526         match_scratch operand is highest.
36528 2017-03-13  Martin Liska  <mliska@suse.cz>
36530         PR middle-end/78339
36531         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36532         is a CHKP clone, use original declaration.
36534 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36536         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36537         (arc_conditional_register_usage): Use a different allocation order
36538         when optimizing for size.
36539         * common/config/arc/arc-common.c (arc_option_optimization_table):
36540         Section anchors default on when optimizing for size.
36542 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36544         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36546 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36548         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36549         * config/arc/arc.md (cpu_facility): Add cd variant.
36550         (*movqi_insn): Add code density variant.
36551         (*movhi_insn): Likewise.
36552         (*movqi_insn): Likewise.
36553         (*addsi3_mixed): Likewise.
36554         (subsi3_insn): Likewise.
36556 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36558         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36560 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36562         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36563         expressions with MINUS and UNARY ops.
36565 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36567         PR target/79911
36568         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36569         Rename to...
36570         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36571         between vec_select and vector argument.
36572         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36573         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36574         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36575         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36576         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36577         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36579 2017-03-13  Richard Biener  <rguenther@suse.de>
36581         PR other/79991
36582         * params.def (vect-max-peeling-for-alignment): Fix typo.
36584 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36586         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36587         issue that only occurred with binutils below 2.18.
36589 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36591         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36592         refer to binutils 2.11/2.12 minimum.
36594 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36596         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36597         ftp.kernel.org and simplify binutils requirement.
36599 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36601         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36602         optimization.
36603         (Optimize Options): Ditto.  Also remove redundancy.
36605 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36607         PR translation/79848
36608         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36609         "%qs".
36610         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36611         to G_ to avoid double translation.
36613 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36615         PR translation/79923
36616         * auto-profile.c (get_combined_location): Convert leading
36617         character of diagnostics to lower case and remove trailing period.
36618         (read_profile): Likewise for various diagnostics.
36619         * config/arm/arm.c (arm_option_override): Remove trailing period
36620         from various diagnostics.
36621         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36622         (msp430_expand_delay_cycles): Likewise.
36624 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36626         PR target/79925
36627         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36628         full command-line argument, rather than just "str".
36629         (aarch64_validate_march): Likewise.
36630         (aarch64_validate_mtune): Likewise.
36632 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36634         PR rtl-optimization/78911
36635         * lra-assigns.c (must_not_spill_p): New function.
36636         (spill_for): Use it.
36638 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36640         PR tree-optimization/79981
36641         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36642         ATOMIC_COMPARE_EXCHANGE ifn result.
36643         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36644         IFN_ATOMIC_COMPARE_EXCHANGE.
36646 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36648         PR driver/79875
36649         * opts.c (parse_sanitizer_options): Add missing question mark to
36650         "did you mean" message.
36652 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36654         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36655         built-in.
36656         (VMULEUH_UNS): Likewise.
36657         (VMULOUB_UNS): Likewise.
36658         (VMULOUH_UNS): Likewise.
36659         * config/rs6000/rs6000.c (builtin_function_type): Remove
36660         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36662 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36664         PR bootstrap/79952
36665         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36666         x with result of extra_parsing_for_operand_code_0.
36667         (function_reader::extra_parsing_for_operand_code_0): Convert
36668         return type from void to rtx, returning x.  When reading
36669         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36670         larger size containing struct block_symbol.
36672 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36674         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36675         -mfloat128-hardware without -m64.
36677 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36679         PR target/79941
36680         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36681         entries to the case statement that marks unsigned arguments to
36682         overloaded functions.
36684 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36686         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36687         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36689 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36691         PR target/79907
36692         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36693         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36695 2017-03-10  Martin Liska  <mliska@suse.cz>
36697         PR target/65705
36698         PR target/69804
36699         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36700         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36701         FIELD != NULL.
36703 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36705         * tree-switch-conversion (array_value_type): Start by resetting
36706         candidate type to it's main variant.
36708 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36710         PR rtl-optimization/79909
36711         * combine.c (try_combine): Use simplify_replace_rtx on individual
36712         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36713         of the whole CALL_INSN_FUNCTION_USAGE.
36715         PR tree-optimization/79972
36716         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36717         get_range_info on SSA_NAMEs.  Formatting fixes.
36719 2017-03-10  Richard Biener  <rguenther@suse.de>
36720             Jakub Jelinek  <jakub@redhat.com>
36722         PR tree-optimization/77975
36723         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36724         edge to be constant.
36725         (get_val_for): For constant x return it.  Formatting fix.
36726         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36727         would use the same bases as the current one.
36729 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36731         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36732         instead of vec_select for V1TImode.
36733         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36734         longer needed.
36735         (VSX_LE_128): Add V1TI to this mode iterator.
36736         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36737         (*vsx_le_perm_store_<mode>): Likewise.
36738         (pre-reload splitter for VSX stores): Likewise.
36739         (post-reload splitter for VSX stores): Likewise.
36740         (*vsx_xxpermdi2_le_<mode>): Likewise.
36741         (*vsx_lxvd2x2_le_<mode>): Likewise.
36742         (*vsx_stxvd2x2_le_<mode>): Likewise.
36744 2017-03-09  Michael Eager  <eager@eagercon.com>
36746         Correct failures with --enable-checking=yes,rtl.
36748         * config/microblaze/microblaze.c (microblaze_expand_shift):
36749         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36750         test for const0_rtx.
36751         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36752         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36754 2017-03-09  Richard Biener  <rguenther@suse.de>
36756         PR tree-optimization/79977
36757         * graphite-scop-detection.c (scop_detection::merge_sese):
36758         Handle the case of extra exits to blocks dominating the entry.
36760 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36762         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36763         Document rdynamic.
36765 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36767         PR rtl-optimization/79949
36768         * lra-constraints.c (process_alt_operands): Check memory when
36769         trying to predict a cycle.  Print about the overall increase.
36771 2017-03-09  Richard Biener  <rguenther@suse.de>
36773         PR middle-end/79971
36774         * gimple-expr.c (useless_type_conversion_p): Preserve
36775         TYPE_SATURATING for fixed-point types.
36777 2017-03-09  Richard Biener  <rguenther@suse.de>
36779         PR ipa/79970
36780         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36781         alignment of BLKmode params.
36783 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36785         PR target/79913
36786         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36787         (VALL_NO_V2Q): Likewise.
36788         (VDQF_DF): Delete.
36789         * config/aarch64/aarch64-simd.md
36790         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36791         iterator.
36792         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36793         VALL_NO_V2Q mode iterator.
36794         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36796 2017-03-09  Martin Liska  <mliska@suse.cz>
36798         PR tree-optimization/79631
36799         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36800         tree_int_cst_sign_bit just for INTEGER constants.
36802 2017-03-09  Martin Liska  <mliska@suse.cz>
36804         PR target/65705
36805         PR target/69804
36806         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36807         sanitizers.
36809 2017-03-09  Marek Polacek  <polacek@redhat.com>
36811         PR c++/79672
36812         * tree.c (inchash::add_expr): Handle TREE_VEC.
36814 2017-03-09  Martin Liska  <mliska@suse.cz>
36816         PR ipa/79764
36817         (chkp_narrow_size_and_offset): New function.
36818         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36819         (void chkp_parse_bit_field_ref): New function.
36820         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36821         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36823 2017-03-09  Martin Liska  <mliska@suse.cz>
36825         PR ipa/79761
36826         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36827         (chkp_find_bounds_1): Remove gcc_unreachable.
36829 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36831         PR sanitizer/79944
36832         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36833         BUILT_IN_SYNC*, determine the access type from the size suffix and
36834         always build a MEM_REF with that type.  Handle forgotten
36835         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36837         PR target/79932
36838         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36839         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36840         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36841         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36842         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36843         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36844         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36845         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36846         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36847         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36848         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36849         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36850         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36851         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36852         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36853         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36854         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36855         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36856         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36857         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36858         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36859         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36860         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36861         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36862         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
36863         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
36864         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
36865         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
36866         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
36867         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
36868         definitions outside of __OPTIMIZE__ guarded section.
36870         PR target/79932
36871         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
36872         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
36873         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
36874         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
36875         guarded section.
36877 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36879         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
36880         ("vfenez<mode>"): Add missing constraints.
36882 2017-03-08  Martin Sebor  <msebor@redhat.com>
36884         PR target/79928
36885         * config/nds32/nds32.c (nds32_option_override):
36886         Fix misspelled diagnostic.
36888 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36890         PR c/79940
36891         * gimplify.c (gimplify_omp_for): Replace index var in outer
36892         taskloop statement with an artificial variable and add
36893         OMP_CLAUSE_PRIVATE clause for it.
36895 2017-03-08  Richard Biener  <rguenther@suse.de>
36897         PR tree-optimization/79955
36898         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
36899         for accesses that are completely outside of the variable.
36901 2017-03-08  Andrew Haley  <aph@redhat.com>
36903         PR tree-optimization/79943
36904         * tree-ssa-loop-split.c (compute_new_first_bound): When
36905         calculating the new upper bound, (END-BEG) should be added, not
36906         subtracted.
36908 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36910         * config/avr/avr.md (setmemhi): Make sure match_dup
36911         operand number comes before match_scratch.
36913 2017-03-08  Richard Biener  <rguenther@suse.de>
36915         PR tree-optimization/79920
36916         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
36917         with ncopies == 1 to ...
36918         (vect_transform_slp_perm_load): ... here.  Properly compute
36919         all element loads by iterating VF times over the group.  Do
36920         not handle ncopies (computed in a broken way) in
36921         vect_create_mask_and_perm.
36923 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36925         PR sanitizer/79904
36926         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
36927         is a uniform vector, use uniform_vector_p return value instead of
36928         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
36930 2017-03-07  Marek Polacek  <polacek@redhat.com>
36932         PR middle-end/79809
36933         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
36934         (alloca_call_type): Likewise.
36936 2017-03-07  Martin Liska  <mliska@suse.cz>
36938         * gcov.c (process_args): Put comment to correct location.
36940 2017-03-07  Martin Liska  <mliska@suse.cz>
36942         PR middle-end/68270
36943         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
36944         Use array_at_struct_end_p instead of DECL_CHAIN (field).
36945         (chkp_narrow_bounds_for_field): Likewise.
36946         (chkp_parse_array_and_component_ref): Pass one more argument to
36947         call.
36949 2017-03-07  Richard Biener  <rguenther@suse.de>
36951         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
36952         preheaders.
36954 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
36956         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
36957         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
36959 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36961         PR c/79855
36962         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
36963         to end of description.
36964         (PARAM_MAX_STORES_TO_MERGE): Likewise.
36966 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
36968         PR rtl-optimization/79901
36969         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
36970         ...
36971         (*avx512f_<code><mode>3<mask_name>): ... this.
36972         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
36973         iterator instead of VI8_AVX2_AVX512BW.
36975         PR rtl-optimization/79901
36976         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
36977         min/max expander, expand it using expand_vec_cond_expr.
36979         PR sanitizer/79897
36980         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
36981         temporary.
36983 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
36985         PR c++/79821
36986         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
36987         to void * for PCH reasons.
36988         * dwarf2out.c (output_loc_operands, output_die): Cast
36989         v.val_vec.array to unsigned char *.
36991 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
36993         PR target/77850
36994         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
36995         vector types.
36997 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
36999         PR rtl-optimization/79571
37000         * lra-constraints.c (process_alt_operands): Calculate static
37001         reject and subtract it from overall when only addresses will be
37002         reloaded.
37004 2017-03-06  Julia Koval  <julia.koval@intel.com>
37006         PR target/79793
37007         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37008         incoming stack boundary to 128 for 64-bit targets.
37010 2017-03-06  Richard Biener  <rguenther@suse.de>
37012         PR tree-optimization/79894
37013         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37014         to NULL after folding it.
37016 2017-03-06  Richard Biener  <rguenther@suse.de>
37018         PR tree-optimization/79824
37019         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37020         check disabling peeling for gaps.
37022 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37024         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37025         attributes): Document gettimeofday.
37027 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37029         * config/s390/s390.c (s390_option_override_internal): Set
37030         PARAM_MIN_VECT_LOOP_BOUND
37032 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37034         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37035         * config/s390/s390.md: Likewise.
37037 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37039         PR target/79812
37040         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37041         (<avx2_avx512>_perm<mode>): Rename to ...
37042         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37043         of VI8F_256_512.
37044         (<avx512>_perm<mode>_mask): Rename to ...
37045         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37046         of VI8F_256_512.
37047         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37048         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37049         instead of VI8F_256_512.
37050         (avx512f_perm<mode>): New define_expand.
37051         (avx512f_perm<mode>_mask): Likewise.
37052         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37053         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37055 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37057         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37058         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37059         if_then_else.
37060         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37062 2017-03-06  Martin Liska  <mliska@suse.cz>
37064         PR sanitize/79783
37065         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37066         when having a SSA NAME w/o VAR_DECL assigned to it.
37068 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37070         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37071         msa_dpsub_<su>_d): Fix MODE for vec_select.
37073 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37075         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37076         argument.
37077         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37079 2017-03-06  Richard Biener  <rguenther@suse.de>
37081         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37082         * plugin.c (register_plugin_info): Likewise.
37083         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37085 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37087         * config/i386/sse.md (sse_storehps, sse_storelps,
37088         avx_<castmode><avxsizesuffix>_<castmode>,
37089         avx512f_<castmode><avxsizesuffix>_<castmode>,
37090         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37091         in condition that at least one operand is not a MEM.
37093 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37095         PR middle-end/79805
37096         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37097         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37098         ECF_NOTHROW.
37099         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37100         gimple_call_nothrow_p flag based on whether original builtin can throw.
37101         If it can, emit following stmts on the fallthrough edge.
37102         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37103         don't create new bb if inserting just debug stmts on the edge, try to
37104         insert them on the fallthru bb or just reset debug stmts.
37106 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37108         PR target/43763
37109         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37110         restore recog_data (including the operand rtxes inside it) around
37111         the call to get_insn_template.
37113 2017-03-03  Martin Sebor  <msebor@redhat.com>
37115         PR tree-optimization/79699
37116         * context.c (context::~context): Free MPFR caches to avoid
37117         a memory leak on program exit.
37119 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37121         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37122         Use wide_int::ulow () instead of .elt (0).
37124 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37126         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37127         (*pushxf): Limit oF constraint to 32bit targets and add oC
37128         constraint for 64bit targets.
37129         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37130         (*pushdf): Change rmF constraint to rmC.
37132 2017-03-03  Martin Liska  <mliska@suse.cz>
37134         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37135         Remove unused variable.
37137 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37139         PR target/79807
37140         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37141         is a memory operand, increase num_memory.
37142         (ix86_expand_args_builtin): Likewise.
37144 2017-03-03  Jan Hubicka  <jh@suse.cz>
37146         PR lto/79760
37147         * ipa-devirt.c (maybe_record_node): Properly handle
37148         __cxa_pure_virtual visibility.
37150 2017-03-03  Martin Liska  <mliska@suse.cz>
37152         PR tree-optimization/79803
37153         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37154         assert.
37155         (pass_loop_prefetch::execute): Disabled optimization if an
37156         assumption about L1 cache size is not met.
37158 2017-03-03  Martin Liska  <mliska@suse.cz>
37160         PR rtl-optimization/79574
37161         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37162         (hash_scan_set): Likewise.
37163         (dump_hash_table): Likewise.
37164         (hoist_code): Likewise.
37166 2017-03-03  Richard Biener  <rguenther@suse.de>
37168         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37169         in place of uhigh ().
37170         (fixed_convert_from_real): Likewise.
37172 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37174         PR target/79514
37175         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37177 2017-03-03  Richard Biener  <rguenther@suse.de>
37179         PR middle-end/79818
37180         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37181         TYPE_OVERFLOW_UNDEFINED check.
37183 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37185         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37186         numbers.
37187         (vector_ae_<mode>_p): Likewise.
37188         (vector_nez_<mode>_p): Likewise.
37189         (vector_ne_v2di_p): Likewise.
37190         (vector_ae_v2di_p): Likewise.
37191         (vector_ne_<mode>_p): Likewise.
37192         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37193         numbers.
37194         (vsx_tsqrt<mode>2_fe): Likewise.
37196 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37198         PR target/79514
37199         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37201 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37203         PR rtl-optimization/79780
37204         * cprop.c (one_cprop_pass): When second and further conditional trap
37205         in a single basic block is turned into an unconditional trap, turn it
37206         into a deleted note to avoid RTL verification failures.
37208 2017-03-02  Richard Biener  <rguenther@suse.de>
37210         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37212 2017-03-02  Richard Biener  <rguenther@suse.de>
37214         PR tree-optimization/79345
37215         PR c++/42000
37216         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37217         param and abort the walk, returning -1 if it is hit.
37218         (walk_aliased_vdefs): Take a limit param and pass it on.
37219         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37220         defaulting to 0 and return a signed int.
37221         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37222         (check_defs): New helper.
37223         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37224         about uninitialized memory.
37225         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37226         bogus uninitialized warning.
37227         (fixed_convert_from_real): Likewise.
37229 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37231         PR tree-optimization/66768
37232         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37233         iv_use if base object can't be determined.
37235 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37237         PR tree-optimization/79345
37238         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37239         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37240         (get_pattern_stats): Initialize it.
37241         * genemit.c (gen_expand): Verify match_scratch numbers come after
37242         match_operand/match_dup numbers.
37243         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37244         match_scratch numbers.
37245         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37246         Likewise.
37247         * config/s390/s390.md (trunctdsd2): Likewise.
37249 2017-03-02  Richard Biener  <rguenther@suse.de>
37251         * wide-int.h (wide_int_storage::operator=): Implement in terms
37252         of wi::copy.
37254 2017-03-02  Richard Biener  <rguenther@suse.de>
37256         PR tree-optimization/79777
37257         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37258         the to insert expression to sth existing.
37260 2017-03-01  Martin Sebor  <msebor@redhat.com>
37262         PR middle-end/79692
37263         * gimple-ssa-sprintf.c
37264         (directive::known_width_and_precision): New function.
37265         (format_integer): Use it.
37266         (get_mpfr_format_length): Consider the full range of precision
37267         when computing %g output with the # flag.  Set the likely byte
37268         count to 3 rather than 1 when precision is indeterminate.
37269         (format_floating): Correct the lower bound of precision.
37271 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37273         * doc/invoke.texi: Document default code model for 64-bit Linux.
37275 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37277         PR target/79752
37278         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37279         udiv rather than div since input pattern is unsigned.
37281 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37283         * config/i386/i386.c (print_reg): Warn for values of
37284         unsupported size in integer register.
37286 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37288         PR target/79439
37289         * config/rs6000/predicates.md (current_file_function_operand): Do
37290         not allow self calls to be local if the function is replaceable.
37292 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37294         PR target/79395
37295         * config/rs6000/altivec.h (vec_ctz and others): Change the
37296         preprocessor macro that controls conditional compilation from
37297         _ARCH_PWR9 to __POWER9_VECTOR__.
37298         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37299         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37300         control (instead of _ARCH_PWR9 control) so that template
37301         definition uses power9-specific function.
37302         (vec_any_eq): Likewise.
37303         (vec_all_ne): Change macro definition to use a power9-specific
37304         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37305         _ARCH_PWR9 control).
37306         (vec_any_eq) Likewise.
37307         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37308         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37309         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37310         support for xvcmpnedp instruction.
37311         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37312         macro expansion so that Power9 implementation of vec_all_ne does
37313         not use the AltiVec predicate framework.
37314         (VCMPNEH_P): Likewise.
37315         (VCMPNEW_P): Likewise.
37316         (VCMPNED_P): Likewise.
37317         (VCMPNEFP_P): Likewise.
37318         (VCMPNEDP_P): Likewise.
37319         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37320         implementation of vec_any_eq to not use AltiVec predicate
37321         framework.
37322         (VCMPAEH_P): Likewise.
37323         (VCMPAEW_P): Likewise.
37324         (VCMPAED_P): Likewise.
37325         (VCMPAEFP_P): Likewise.
37326         (VCMPAEDP_P): Likewise.
37327         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37328         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37329         not use the AltiVec predicate framework.
37330         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37331         of vec_any_eq to not use AltiVec predicate framework.
37332         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37333         support for predefined __POWER9_VECTOR__ macro to indicate that
37334         Power9 instruction selection is enabled.
37335         (altivec_overloaded_builtins): Remove extraneous
37336         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37337         function argument types RS6000_BTI_bool_V16QI and
37338         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37339         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37340         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37341         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37342         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37343         Power9 for implementations of vec_cmpne.  Change the signature for
37344         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37345         (representing vec_all_ne) to remove the previously described first
37346         argument of type RS6000_BTI_INTSI, as this was an artifact of
37347         reliance on the AltiVec predicate framework, which is no longer
37348         used in the implementation of these functions.  Add
37349         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37350         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37351         since, unlike the AltiVec predicate framework implementation, we
37352         do not share function descriptors between vec_alle and vec_anyeq.
37353         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37354         set of modes that receive special treatment even when
37355         TARGET_P9_VECTOR is true.  The special treatment emits code that
37356         does not depend on Power9 instructions.
37357         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37358         define_expand to not rely on AltiVec predicate framework.
37359         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37360         function.
37361         (vector_ne_v2di_p): Change this define_expand to not rely on
37362         AltiVec predicate framework.
37363         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37364         function.
37365         (vector_ne_<mode>_p): Change this define_expand to not rely on
37366         AltiVec predicate framework.
37367         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37368         function.
37369         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37370         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37371         define_insn pattern.
37372         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37373         define_insn pattern because the xvcmpne<VSs>. instruction is not
37374         supported.
37375         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37376         instruction is not supported.
37378 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37380         * config/nvptx/nvptx.c: Include intl.h.
37382 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37384         PR lto/78140
37385         * ipa-prop.h (ipa_bits): Removed field known.
37386         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37387         to pointers.  Adjusted their comments to warn about their sharing.
37388         (ipcp_transformation_summary): Change bits to a vector of pointers.
37389         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37390         (ipa_get_ipa_bits_for_value): Declare.
37391         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37392         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37393         (ipa_bits_hash_table): Likewise.
37394         (ipa_vr_ggc_hash_traits): Likewise.
37395         (ipa_vr_hash_table): Likewise.
37396         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37397         being pointers and vr_known being removed.
37398         (ipa_set_jf_unknown): Likewise.
37399         (ipa_get_ipa_bits_for_value): New function.
37400         (ipa_set_jfunc_bits): Likewise.
37401         (ipa_get_value_range): New overloaded functions.
37402         (ipa_set_jfunc_vr): Likewise.
37403         (ipa_compute_jump_functions_for_edge): Use the above functions to
37404         construct bits and vr parts of jump functions.
37405         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37406         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37407         exist.
37408         (ipcp_grow_transformations_if_necessary): Also allocate
37409         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37410         exist.
37411         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37412         them.  Fix too long lines.
37413         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37414         vr_known being removed.
37415         (ipa_read_jump_function): Use new setter functions to construct bits
37416         and vr parts of jump functions or set them to NULL.
37417         (write_ipcp_transformation_info): Adjust for bits being pointers.
37418         (read_ipcp_transformation_info): Likewise.
37419         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37420         space.
37421         Include gt-ipa-prop.h.
37422         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37423         being pointers.
37424         (ipcp_store_bits_results): Likewise.
37425         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37426         Do not write to existing jump functions but use a temporary instead.
37428 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37430         PR c++/79681
37431         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37432         attempt to use its first operand as BIT_FIELD_REF base.
37434 2017-03-01  Richard Biener  <rguenther@suse.de>
37436         PR middle-end/79721
37437         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37438         interpolating formula in wrapping arithmetic.
37439         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37441 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37443         PR tree-optimization/79734
37444         * tree-vect-generic.c (expand_vector_condition): Optimize
37445         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37446         Handle VEC_COND_EXPR where comparison has different inner width from
37447         type's inner width.
37449 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37451         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37452         markup, and similar issues.  Remove @opindex entries for things
37453         that aren't options.  Add missing -mmpy-option entries.
37455 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37457         PR tree-optimization/79737
37458         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37459         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37460         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37461         instead of byte_size.  Formatting fix.
37462         (shift_bytes_in_array_right): Formatting fix.
37464 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37466         PR target/79749
37467         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37468         condition on optimize for the leaf function test.
37470 2017-02-28  Martin Liska  <mliska@suse.cz>
37472         PR lto/79625
37473         * read-rtl-function.c (function_reader::handle_unknown_directive):
37474         Bail out when one uses -flto.
37476 2017-02-28  Martin Liska  <mliska@suse.cz>
37478         * common.opt: Replace space with tabular for options of <number>
37479         type.
37480         * config/i386/i386.opt: Show <number> value for
37481         -mlarge-data-threshold.
37482         * opts.c (print_filtered_help): Do not display number in hexadecimal
37483         format.
37485 2017-02-28  Martin Liska  <mliska@suse.cz>
37487         * common.opt: Fix --help=option -Q for options which are of
37488         an enum type.
37490 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37492         * config/i386/i386.c (print_reg): Error out for values
37493         of 8-bit size in invalid integer register.
37495 2017-02-28  Martin Sebor  <msebor@redhat.com>
37497         PR tree-optimization/79691
37498         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37500 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37502         PR target/79729
37503         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37504         gcc_unreachable with output_operand_lossage.
37506 2017-02-28  Richard Biener  <rguenther@suse.de>
37508         PR tree-optimization/79740
37509         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37510         inserts.
37511         (visit_nary_op): Insert the nary into the hashtable if we
37512         pattern-matched sth.
37513         * tree-ssa-pre.c (eliminate_insert): Robustify.
37515 2017-02-28  Richard Biener  <rguenther@suse.de>
37517         PR middle-end/79731
37518         * fold-const.c (decode_field_reference): Reject out-of-bound
37519         accesses.
37521 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37523         * config/i386/i386.c: Include intl.h.
37524         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37525         instead of just cond ? "..." : "...".
37526         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37527         * coverage.c (read_counts_file): Likewise.
37528         * omp-offload.c: Include intl.h.
37529         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37530         of just cond ? "..." : "...".
37531         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37532         of just cond ? "..." : "...".
37534 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37536         PR target/79742
37537         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37538         entry, if present.
37539         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37540         'tune for' CPU name.
37541         * config/arm/arm-cpu-data.h: Regenerated.
37543 2017-02-28  Richard Biener  <rguenther@suse.de>
37545         PR tree-optimization/79732
37546         * tree-inline.c (expand_call_inline): Do not shadow var.
37548 2017-02-28  Richard Biener  <rguenther@suse.de>
37550         PR tree-optimization/79723
37551         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37552         address-space properly.
37554 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37556         * doc/optinfo.texi (Optimization groups): Fix option used for
37557         OPTGROUP_ALL.
37558         * doc/invoke.texi (-fopt-info): Document "omp".
37559         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37560         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37561         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37562         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37563         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37565         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37566         all users.
37567         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37568         OPTGROUP_OMP with "omp".
37570 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37572         PR target/79544
37573         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37574         for arithmetic shift of unsigned V2DI.
37576 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37578         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37579         arc/linux.h headers.
37580         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37581         (LINK_SPEC): Likewise.
37582         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37583         (EXTRA_SPECS): Likewise.
37584         (STARTFILE_SPEC): Likewise.
37585         (ENDFILE_SPEC): Likewise.
37586         (LIB_SPEC): Likewise.
37587         (TARGET_SDATA_DEFAULT): Likewise.
37588         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37589         (MULTILIB_DEFAULTS): Likewise.
37590         (DWARF2_UNWIND_INFO): Likewise.
37591         * config/arc/big.h: New file.
37592         * config/arc/elf.h: Likewise.
37593         * config/arc/linux.h: Likewise.
37594         * config/arc/t-uClibc: Remove.
37596 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37598         PR tree-optimization/77536
37599         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37600         (tree_transform_and_unroll_loop): Use above function to compute the
37601         estimated niter of unrolled loop and use it when scaling profile.
37602         Also use count info rather than frequency if it's non-zero.
37603         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37604         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37605         (vect_transform_loop): Call above function.
37607 2017-02-27  Richard Biener  <rguenther@suse.de>
37609         PR tree-optimization/45397
37610         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37611         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37612         (visit_nary_op): Add pattern matching for CSEing sign-changed
37613         or truncated operations with wider ones.
37615 2017-02-27  Richard Biener  <rguenther@suse.de>
37617         PR tree-optimization/79690
37618         * tree-vect-stmts.c (vectorizable_store): Use vector type
37619         built from the DR with address-space.
37621 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37623         * doc/invoke.texi (Optimize Options): Refine the description
37624         of asan-use-after-return.
37626 2017-02-25  Alan Modra  <amodra@gmail.com>
37628         PR rtl-optimization/79584
37629         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37630         base, not ad->base_term, the reg within base.  Remove assertion
37631         that ad->base == ad->base_term.  Replace gen_int_mode using
37632         bogus mode with const0_rtx.
37634 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37636         PR middle-end/79396
37637         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37638         FMA_EXPR like tcc_binary or tcc_unary.
37640         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37642         PR debug/77589
37643         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37644         bitfield.
37645         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37646         (output_loc_operands): Handle DW_OP_call_ref and
37647         DW_OP_GNU_variable_value.
37648         (struct variable_value_struct): New type.
37649         (struct variable_value_hasher): Likewise.
37650         (variable_value_hash): New variable.
37651         (string_types): Remove.
37652         (copy_loc_descr): New function.
37653         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37654         (prepend_loc_descr_to_each): New function.
37655         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37656         instead of add_loc_descr_to_each if the first argument is single
37657         location list and the second has multiple.
37658         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37659         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37660         when looking for variable value which doesn't have other location info.
37661         (loc_list_from_tree): Formatting fix.
37662         (gen_array_type_die): Simplify DW_AT_string_length handling.
37663         (adjust_string_types): Remove.
37664         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37665         string_types.  Call resolve_variable_values.
37666         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37667         (resolve_addr_in_expr): Likewise.  Add A argument.
37668         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37669         original expression being DW_OP_GNU_variable_value with optionally
37670         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37671         optionally after it.
37672         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37673         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37674         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37675         (variable_value_hasher::hash, variable_value_hasher::equal): New
37676         methods.
37677         (resolve_variable_value_in_expr, resolve_variable_value,
37678         resolve_variable_values, note_variable_value_in_expr,
37679         note_variable_value): New functions.
37680         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37681         DIEs.
37683 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37685         PR c/79677
37686         * opts.h (handle_generated_option): Add GENERATED_P argument.
37687         * opts-common.c (handle_option): Adjust function comment.
37688         (handle_generated_option): Add GENERATED_P argument, pass it to
37689         handle_option.
37690         (control_warning_option): Pass false to handle_generated_option
37691         GENERATED_P.
37692         * opts.c (maybe_default_option): Pass true to handle_generated_option
37693         GENERATED_P.
37694         * optc-gen.awk: Likewise.
37696 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37698         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37699         a REG, look at the REG it is a SUBREG of.
37700         (splitter for cmpeqsi_t): Ditto.
37702 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37704         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37705         the special USEs with the pattern of the insn, not the insn itself.
37707 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37709         PR target/79473
37710         * doc/invoke.texi: Document -mload-store-pairs.
37712 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37713             Sandra Loosemore  <sandra@codesourcery.com>
37715         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37716         argument isn't a CONST_INT.
37717         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37718         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37719         (nios2_validate_compare): Bypass alternate compare logic if *op2
37720         is not a CONST_INT.
37721         (ldstwm_operation_p): Return false if first_base is not a REG or
37722         if first_offset is not a CONST_INT.
37724 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37726         * config/cris/cris.md: Use correct operand in a define_peephole2.
37728 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37730         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37732 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37734         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37735         this_insn if it is an INSN or JUMP_INSN.
37736         (force_offsettable): Look at base, not at addr.
37737         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37738         on things that aren't necessarily CONST_INTs.
37740 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37742         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37743         -mfpmath=sse is the default also for x86-32 targets with SSE2
37744         instruction set when @option{-ffast-math} is enabled
37746 2017-02-24  Jeff Law  <law@redhat.com>
37748         PR rtl-optimizatoin/79286
37749         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37750         dominance test.
37752 2017-02-24  Richard Biener  <rguenther@suse.de>
37754         PR tree-optimization/79389
37755         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37756         debug insns.
37758 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37760         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37761         function comment to reflect reality.
37762         (loop_exits_before_overflow): Fix typo in function description.
37764 2017-02-24  Richard Biener  <rguenther@suse.de>
37766         PR tree-optimization/79389
37767         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37768         properly that a threading opportunity exists.  Detect conditional
37769         copy/constant propagation opportunities.
37771 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37773         * config/visium/visium.md (type): Add trap.
37774         (b): New mode attribute.
37775         (*btst): Rename into...
37776         (*btst<mode>): ...this and adjust.
37777         (*cbranchsi4_btst_insn): Rename into...
37778         (*cbranch<mode>4_btst_insn): ...this and adjust.
37779         (trap): New define_insn.
37781 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37783         PR tree-optimization/79389
37784         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37785         (noce_reversed_cond_code): New function.
37786         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37787         reversed_comparison_code.  Formatting fix.
37788         (noce_try_store_flag): Test rev_cond != NULL in addition to
37789         reversed_comparison_code.
37790         (noce_try_store_flag_constants): Likewise.
37791         (noce_try_store_flag_mask): Likewise.
37792         (noce_try_addcc): Use rev_cond if non-NULL instead of
37793         reversed_comparison_code.
37794         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37795         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37796         (noce_find_if_block): Initialize rev_cond.
37797         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37798         instead of false as last argument never attempt to reverse it
37799         afterwards.
37801 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37803         PR tree-optimization/79663
37804         * tree-predcom.c (combine_chains): Process refs in reverse order
37805         only for ZERO length chains, and add explaining comment.
37807 2017-02-23  Jeff Law  <law@redhat.com>
37809         PR tree-optimization/79578
37810         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37811         in call to operand_equal_p.
37813 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37815         PR target/71017
37816         * config/i386/cpuid.h: Fix another undefined behavior.
37818 2017-02-23  Richard Biener  <rguenther@suse.de>
37820         PR tree-optimization/79683
37821         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37822         vector types for data-refs.
37824 2017-02-23  Martin Liska  <mliska@suse.cz>
37826         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37828 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37830         PR middle-end/79665
37831         * internal-fn.c (get_range_pos_neg): Moved to ...
37832         * tree.c (get_range_pos_neg): ... here.  No longer static.
37833         * tree.h (get_range_pos_neg): New prototype.
37834         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37835         are known to be in between 0 and signed maximum inclusive, try to
37836         expand both unsigned and signed divmod and use the cheaper one from
37837         those.
37839 2017-02-22  Jeff Law  <law@redhat.com>
37841         PR tree-optimization/79578
37842         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37843         to compare base operands.
37845 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37847         PR target/79211
37848         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37849         gpc_reg_operand instead of fpr_reg_operand.
37851 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
37853         * config/mips/mips.c (mips_return_in_memory): Force FP
37854         vector types to be returned in memory for o32 ABI.
37856 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37858         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37859         instead of DW_TAG_member for static data member declarations and don't
37860         set no_linkage_name for static inline data members.
37861         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37862         to DW_TAG_member.
37864 2017-02-22  Martin Liska  <mliska@suse.cz>
37866         * doc/invoke.texi: Replace inequality signs with square brackets
37867         for -Wnormalized.
37869 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
37871         PR target/78660
37872         * lra-constraints.c (simplify_operand_subreg): Handle
37873         WORD_REGISTER_OPERATIONS targets.
37875 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37877         PR target/70465
37878         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
37879         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
37880         elimination by swapping fld*.
37882 2017-02-22  Richard Biener  <rguenther@suse.de>
37884         PR tree-optimization/79673
37885         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
37886         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
37887         irrelevant address-space qualifiers and avoiding a
37888         ADDR_SPACE_CONVERT_EXPR from fold_convert.
37890 2017-02-22  Richard Biener  <rguenther@suse.de>
37892         PR tree-optimization/79666
37893         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
37894         to not symbolically negate if that may introduce undefined
37895         overflow.
37897 2017-02-22  Martin Liska  <mliska@suse.cz>
37899         PR lto/79587
37900         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
37901         * data-streamer-out.c (streamer_write_gcov_count_stream):
37902         Likewise.
37903         * value-prof.c (stream_out_histogram_value): Make assert more
37904         precise based on type of counter.
37906 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
37908         PR target/79593
37909         * config/i386/i386.md (standard_x87sse_constant_load splitter):
37910         Use nonimmediate_operand instead of memory_operand for operand 1.
37911         (float-extend standard_x87sse_constant_load splitter): Ditto.
37913 2017-02-21  Jeff Law  <law@redhat.com>
37915         PR tree-optimization/79621
37916         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
37917         blocks with edges to themselves.
37919 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37921         PR target/79633
37922         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
37923         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
37924         Use gimple_call_builtin_p.
37926         PR target/79570
37927         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
37928         on temporarily removed DEBUG_INSNs.
37930         PR tree-optimization/79649
37931         * tree-loop-distribution.c (classify_partition): Give up on
37932         non-generic address space loads/stores.
37934 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
37936         * doc/loop.texi (Loop manipulation): Remove nonexistent
37937         tree_ssa_loop_version from the documentation.
37938         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
37940 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37942         PR target/79494
37943         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
37944         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
37945         * config/rs6000/rs6000.c: Include except.h.
37946         (rs6000_expand_split_stack_prologue): Call
37947         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
37949 2017-02-21  Martin Jambor  <mjambor@suse.cz>
37951         PR lto/79579
37952         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
37953         have been analyzed.
37955 2017-02-21  Martin Jambor  <mjambor@suse.cz>
37957         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
37958         for backward compatibility only.
37959         * doc/invoke.texi (Option Summary): Remove all references to
37960         -fipa-cp-alignment.
37962 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
37964         PR target/78660
37965         Revert:
37966         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
37968         * lra-constraints.c (curr_insn_transform): Handle
37969         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
37971 2017-02-21  Martin Liska  <mliska@suse.cz>
37973         * config/i386/i386.opt: Replace -masm-dialect with -masm.
37975 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
37977         PR translation/79638
37978         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
37980 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
37982         PR ada/67205
37983         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
37984         (arm_function_ok_for_sibcall): Return false for an indirect call by
37985         descriptor if all the argument registers are used.
37986         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
37987         alignment of the function.
37989 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37991         PR tree-optimization/61441
37992         * simplify-rtx.c (simplify_const_unary_operation): For
37993         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
37994         the sNaN unmodified.
37996 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
37998         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
37999         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38000         instead of SYSTEM_HEADER_DIR.
38002 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38003             Martin LiÅ¡ka  <mliska@suse.cz>
38005         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38006         Fix typos and grammar, use active voice, and clarify.
38008 2017-02-20  Marek Polacek  <polacek@redhat.com>
38010         PR middle-end/79537
38011         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38013         PR sanitizer/79558
38014         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38016 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38018         PR target/79568
38019         * config/i386/i386.c (ix86_expand_builtin): Handle
38020         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38021         ix86_builtins_isa[fcode].isa as a requirement of those
38022         flags and any other flag in the bitmask.
38023         (ix86_init_mmx_sse_builtins): Use 0 instead of
38024         ~OPTION_MASK_ISA_64BIT as mask.
38025         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38026         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38027         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38028         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38030 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38032         PR target/78012
38033         * lra-constraints.c (split_reg): Check requested split mode
38034         is supported by the register.
38036 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38038         * lra-constraints.c (simplify_operand_subreg): Remove early
38039         return false.
38041 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38043         PR target/78660
38044         * lra-constraints.c (curr_insn_transform): Tighten condition
38045         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38047 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38049         PR target/78660
38050         * lra-constraints.c (curr_insn_transform): Handle
38051         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38053 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38055         Revert:
38056         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38058         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38060 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38062         PR c++/69523
38063         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38064         description.
38066 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38068         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38069         for FMA_EXPR.
38071 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38073         * final.c (last_columnnum, override_columnnum): New variables.
38074         (final_start_function): Set last_columnnum, pass it to begin_prologue
38075         hook and pass 0 to dwarf2out_begin_prologue.
38076         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38077         to source_line debug hook.
38078         (notice_source_line): Compute last_columnnum and for debug_column_info
38079         return true on column changes.
38080         * debug.h (struct gcc_debug_hooks): Add column argument to
38081         source_line and begin_prologue hooks.
38082         (debug_nothing_int_charstar_int_bool): Remove prototype.
38083         (debug_nothing_int_int_charstar,
38084         debug_nothing_int_int_charstar_int_bool): New prototypes.
38085         (dwarf2out_begin_prologue): Add column argument.
38086         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38087         begin_prologue hooks.
38088         (debug_nothing_int_charstar_int_bool): Remove.
38089         (debug_nothing_int_int_charstar,
38090         debug_nothing_int_int_charstar_int_bool): New functions.
38091         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38092         through to dwarf2out_source_line.
38093         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38094         (dwarf2out_source_line): Add column argument, emit it if requested.
38095         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38096         arguments.
38097         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38098         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38099         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38100         through to dwarf2out_begin_prologue.
38101         (vmsdbgout_source_line): Add column argument, pass it through to
38102         dwarf2out_source_line.
38103         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38104         dbxout_source_line caller.
38105         (dbxout_source_line): Add column argument.
38107         * common.opt (gno-column-info, gcolumn-info): New options.
38108         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38109         (check_die): Also test for multiple DW_AT_decl_column attributes.
38110         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38111         DW_AT_decl_column if requested.
38112         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38113         if requested.
38114         (gen_variable_die): Likewise.
38115         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38116         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38118         PR target/79569
38119         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38120         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38121         (ix86_handle_option): Handle OPT_m3dnowa.
38122         * doc/invoke.texi (-m3dnowa): Document.
38123         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38124         -m3dnowa instead of -m3dnow -march=athlon.
38126         PR target/79559
38127         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38128         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38130 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38132         PR target/79261
38133         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38134         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38135         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38136         generator for vsx_xxpermdi_<mode>_be.
38137         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38138         force big-endian semantics.
38139         (vsx_xxpermdi_<mode>_be): New define_expand with same
38140         implementation as previous version of vsx_xxpermdi_<mode>.
38142 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38144         PR tree-optimization/79327
38145         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38146         variable, its initialization and use.
38148 2017-02-17  Julia Koval  <julia.koval@intel.com>
38150         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38151         (OPTION_MASK_ISA_PKU_UNSET): New.
38152         (ix86_handle_option): Handle -mrdpid.
38153         * config/i386/cpuid.h (bit_RDPID): New.
38154         * config/i386/driver-i386.c (host_detect_local_cpu):
38155         Detect RDPID feature.
38156         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38157         * config/i386/i386-c.c (ix86_target_macros_internal):
38158         Handle RDPID flag.
38159         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38160         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38161         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38162         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38163         * config/i386/i386.md (define_insn "rdpid"): New.
38164         * config/i386/i386.opt Add -mrdpid.
38165         * config/i386/immintrin.h (_rdpid_u32): New.
38167 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38169         PR rtl-optimization/79541
38170         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38171         instead of transforming it into USE.
38173 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38175         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38176         If HONOR_SNANS (SFmode) force the input to a register.
38177         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38178         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38179         an frsp or similar insn.
38181 2017-02-17  Martin Liska  <mliska@suse.cz>
38183         PR rtl-optimization/79577
38184         * params.def (selsched-max-sched-times): Increase minimum to 1.
38186 2017-02-17  Martin Liska  <mliska@suse.cz>
38188         PR rtl-optimization/79574
38189         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38191 2017-02-17  Martin Liska  <mliska@suse.cz>
38193         PR tree-optimization/79529
38194         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38195         ssa_defined_default_def_p to handle cases which are implicitly
38196         defined.
38197         * tree-ssa.c (ssa_defined_default_def_p): New function.
38198         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38199         which are implicitly defined.
38200         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38202 2017-02-17  Richard Biener  <rguenther@suse.de>
38204         PR middle-end/79576
38205         * params.def (max-ssa-name-query-depth): Limit to 10.
38207 2017-02-17  Richard Biener  <rguenther@suse.de>
38209         PR tree-optimization/79552
38210         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38211         default defs.
38213 2017-02-17  Richard Biener  <rguenther@suse.de>
38215         PR bootstrap/79567
38216         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38218 2017-02-17  Marek Polacek  <polacek@redhat.com>
38220         PR middle-end/79536
38221         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38222         (fold_negate_expr): New wrapper.
38224 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38226         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38227         Correct terminology and de-emphasize pre-standard behavior.
38229 2017-02-16  Alan Modra  <amodra@gmail.com>
38231         PR rtl-optimization/79286
38232         * ira.c (def_dominates_uses): New function.
38233         (update_equiv_regs): Don't create an equivalence for insns that
38234         may trap where the register def does not dominate the use.
38236 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38238         PR rtl-optimization/78127
38239         * lra.c (lra): Call lra_eliminate before finish the loop after
38240         lra_constraint.
38242 2017-02-16  Richard Biener  <rguenther@suse.de>
38244         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38245         isl/isl_val.h.
38246         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38247         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38248         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38249         (isl_val_int_from_wi): New function.
38250         (extract_affine_gmp): Rename to ...
38251         (extract_affine_wi): ... this, take a widest_int.
38252         (extract_affine_int): Just wrap extract_affine_wi.
38253         (add_param_constraints): Use isl_val_int_from_wi.
38254         (add_loop_constraints): Likewise, and extract_affine_wi.
38256 2017-02-15  Jeff Law  <law@redhat.com>
38258         PR middle-end/79521
38259         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38260         ira_init_register_move_cost_if_necessary.
38262 2017-02-15  Martin Sebor  <msebor@redhat.com>
38264         PR middle-end/32003
38265         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38266         removed in a prior commit.
38268 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38270         PR tree-optimization/79347
38271         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38272         counters during peeling.
38274 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38276         * Makefile.in (site.exp): Remove "set ISLVER".
38278 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38280         PR target/79487
38281         * real.c (real_from_integer): Call real_convert even for decimal.
38283 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38285         PR target/79421
38286         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38288 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38290         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38291         cores and change the partno/implementer to be correct.
38292         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38293         the 'B" as the implementer.
38294         * config/aarch64/aarch64-tune.md: Regenerate.
38296 2017-02-14  Carl Love  <cel@us.ibm.com>
38298         * config/rs6000/rs6000.c: Add case statement entry to make the
38299         xvcvuxdsp built-in argument unsigned.
38300         * config/rs6000/vsx.md: Fix the source and return operand types so they
38301         match the instruction definitions from the ISA document.  Fix typo
38302         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38303         statement.
38305 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38307         PR target/79282
38308         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38309         member early_clobber_alts.
38310         * lra-lives.c (reg_early_clobber_p): New.
38311         (process_bb_lives): Use it.
38312         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38313         (debug_operand_data): Initialize early_clobber_alts.
38314         (setup_operand_alternative): Set up early_clobber_alts.
38315         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38316         alternatives to new_insn_reg.
38317         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38318         it.
38319         (lra_update_insn_regno_info): Pass the new arg.
38321 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38323         PR middle-end/79505
38324         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38325         (new_oacc_loop_raw): Don't clear already cleared fields.
38327         PR target/79481
38328         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38329         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38330         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38332 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38334         PR target/79495
38335         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38337 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38339         PR target/79498
38340         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38341         the extra instruction to the right place to store 128-bit constant
38342         when needed.
38344 2017-02-14  Martin Sebor  <msebor@redhat.com>
38346         PR middle-end/79448
38347         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38348           warning for strings of unknown length.
38350 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38352         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38354 2017-02-14  Jeff Law  <law@redhat.com>
38356         PR target/79404
38357         * ira-costs.c (scan_one_insn): Initialize register move costs
38358         for pseudos seen in USE/CLOBBER insns.
38360         PR tree-optimization/79095
38361         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38362         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38363         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38364         if the operands are known to be not equal, then the resulting range
38365         is ~[0,0].
38366         (intersect_ranges): If the new range is ~[0,0] and the old range is
38367         wide, then prefer ~[0,0].
38368         * tree-vrp.c (overflow_comparison_p_1): New function.
38369         (overflow_comparison_p): New function.
38370         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38371         if NAME is used in an overflow test.
38372         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38373         overflow check that can be expressed as an equality test, then adjust
38374         ops to be that equality test.
38376 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38378         * config/s390/s390-builtin-types.def: Remove flags argument.
38379         * config/s390/s390.c (s390_init_builtins): Likewise.
38381 2017-02-14  Martin Liska  <mliska@suse.cz>
38383         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38384         vector.  Fix trailing white spaces.
38386 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38388         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38389         HFmode.
38391 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38393         PR rtl-optimization/68664
38394         * config/arm/arm.c (arm_sched_can_speculate_insn):
38395         New function.  Declare prototype.
38396         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38398 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38400         PR rtl-optimization/68664
38401         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38402         New function.
38403         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38405 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38407         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38408         max skip bytes for function, loop and jump.
38410 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38412         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38413         ABS_EXPR for gimple dump.
38415 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38417         PR target/79462
38418         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38420         PR tree-optimization/79408
38421         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38422         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38423         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38424         also if rhs1 is INTEGER_CST.
38426 2017-02-14  Richard Biener  <rguenther@suse.de>
38428         PR middle-end/79432
38429         * tree-into-ssa.c (insert_phi_nodes): When the function can
38430         have abnormal edges rewrite SSA names with broken use-def
38431         dominance out of SSA and register them for PHI insertion.
38433 2017-02-13  Martin Sebor  <msebor@redhat.com>
38435         PR middle-end/79496
38436         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38437         clearing info.nowrite flag when snprintf size argument is a range.
38439 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38441         * cprop.c (cprop_jump): Add missing space in string literal.
38442         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38443         (get_constraint_for_component_ref): Likewise.
38444         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38445         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38446         * lra-constraints.c (process_alt_operands): Likewise.
38447         * ipa-inline.c (inline_small_functions): Likewise.
38448         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38449         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38450         * trans-mem.c (diagnose_tm_1_op): Likewise.
38451         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38452         (grid_parallel_clauses_gridifiable): Likewise.
38454         * config/nvptx/mkoffload.c (process): Add space in between
38455         , and %d.
38457         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38458         "MOD4_SSE_REGS" and "ALL_REGS".
38460         * spellcheck.c (test_data): Add , in between "foo" and "food".
38462 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38464         PR target/79449
38465         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38466         boundary crossing check and subsequent code generation agree.
38468 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38470         * config/aarch64/aarch64.c (has_memory_op): Delete.
38471         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38472         has_memory_op.
38474 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38476         PR rtl-optimization/79388
38477         PR rtl-optimization/79450
38478         * combine.c (distribute_notes): When removing TEM_INSN for which
38479         corresponding dest has last value recorded, invalidate that last
38480         value.
38482 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38484         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38485         of explicit '@'.  Add missing assembly comment marker on branch costs
38486         printout.
38488 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38490         * gengtype-lex.l (<in_struct>): Add '/'.
38492 2017-02-13  Martin Liska  <mliska@suse.cz>
38494         PR c/79471
38495         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38497 2017-02-13  Richard Biener  <rguenther@suse.de>
38499         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38500         Remove.
38501         * configure: Re-generate.
38502         * config.in: Likewise.
38503         * graphite-dependences.c: Simplify as if
38504         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38505         * graphite-isl-ast-to-gimple.c: Likewise.
38506         * graphite-optimize-isl.c: Likewise.
38507         * graphite-poly.c: Likewise.
38508         * graphite-sese-to-poly.c: Likewise.
38509         * graphite.h: Likewise.
38510         * toplev.c: Include isl/version.h and use isl_version () for
38511         printing the ISL version.
38512         * doc/install.texi: Update ISL requirement.
38514 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38516         * doc/standards.texi (Standards): Update reference to
38517         Objective-C 2.0.
38519 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38521         * doc/extend.texi (Named Address Spaces): sourceware.org now
38522         defaults to https.
38523         * doc/install.texi (Binaries): Ditto.
38524         (Specific): Ditto.
38526 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38528         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38529         standard terminology "stringize"/"stringizing" throughout.
38530         * doc/cppinternals.texi: Likewise.
38532 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38534         * doc/extend.texi: Fix some spelling mistakes and typos.
38535         * doc/invoke.texi: Likewise.
38537 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38539         PR ipa/79224
38540         * params.def (inline-min-speedup) Change from 10 to 8.
38542 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38544         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38545         4.5.
38547 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38549         PR ipa/79224
38550         * ipa-inline-analysis.c (get_minimal_bb): New function.
38551         (record_modified): Use it.
38552         (remap_edge_change_prob): Handle also ancestor functions.
38554 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38556         * doc/contrib.texi (Contributors): Remove broken link into
38557         the Mauve CVS repository.
38559 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38561         PR middle-end/79454
38562         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38563         result computation whenever lhs doesn't have vector mode, not
38564         just when it has BLKmode.
38566 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38568         * doc/makefile.texi (profiledbootstrap): Refer to the
38569         installation instructions only in textual form.
38571 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38573         PR target/79295
38574         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38576 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38578         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38579         (Specific): Update mingw-w64 reference.
38580         (Binaries): Ditto.
38581         (Specific): Remove broken link to Renesas RX processor.
38583 2017-02-10  Richard Biener  <rguenther@suse.de>
38585         * toplev.c (process_options): Do not mention obsolete graphite
38586         options when printing sorry message about missing graphite support.
38587         Mention -floop-nest-optimize.
38589 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38591         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38592         (vtst_p16): Likewise.
38593         (vtstq_p8): Likewise.
38594         (vtstq_p16): Likewise.
38595         (vtst_p64): New.
38596         (vtstq_p64): Likewise.
38597         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38598         (vset_lane_p64): New.
38599         (vsetq_lane_p64): New.
38601 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38603         PR tree-optimization/79411
38604         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38605         stmt operands are SSA_NAMEs used in abnormal phis.
38606         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38607         phis.
38609 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38611         PR ipa/70795
38612         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38613         flag if needed.
38615 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38617         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38619 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38621         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38622         to avoid warning.
38624         PR c/79413
38625         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38626         not arbitrary TREE_CONSTANT.
38628         PR c/79431
38629         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38630         "omp declare target link" attribute unless is_global_var.
38631         * omp-offload.c (find_link_var_op): Likewise.
38633 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38634             Chung-Lin Tang  <cltang@codesourcery.com>
38636         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38637         OMP_CLAUSE_TILE.
38638         (gimplify_adjust_omp_clauses): Don't delete TILE.
38639         (gimplify_omp_for): Deal with TILE.
38640         * internal-fn.c (expand_GOACC_TILE): New function.
38641         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38642         (GOACC_TILE): New.
38643         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38644         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38645         element fields.
38646         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38647         avoid DIV for outermost collapse var.
38648         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38649         Remove out of date comments, fix whitespace.
38650         * omp-general.c (omp_extract_for_data): Deal with tiling.
38651         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38652         adjust OLF_DIM_BASE value.
38653         (struct omp_for_data): Add tiling field.
38654         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38655         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38656         for auto loops.  Remove default auto determining, moved to
38657         oacc_loop_fixed_partitions.
38658         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38659         stmts, add e_mask field.
38660         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38661         (oacc_thread_numbers): Use oacc_dim_call.
38662         (oacc_xform_tile): New.
38663         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38664         (finish_oacc_loop): Adjust for ifns vector.
38665         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38666         add case for GOACC_TILE fns.
38667         (oacc_loop_xform_loop): Delete.
38668         (oacc_loop_process): Iterate over call list directly, and add
38669         handling for GOACC_TILE fns.
38670         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38671         dump partitioning.
38672         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38673         vector partitioning to outer loops.  Assign 2 partitions to loops
38674         when available. Add TILE handling.
38675         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38676         (execite_oacc_device_lower): Process GOACC_TILE fns,
38677         ignore unknown specs.
38678         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38679         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38680         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38682 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38684         * configure.ac (ACX_BUGURL): Update.
38685         * configure: Regenerate.
38687 2017-02-09  Richard Biener  <rguenther@suse.de>
38689         PR tree-optimization/69823
38690         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38691         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38693 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38695         * config/arc/arc-c.def: Add __NPS400__ definition.
38696         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38697         (TARGET_NPS400): Define.
38699 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38701         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38702         file.
38703         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38704         pointer, arch_info.
38705         (arc_cpu_types): Fill the arch_info field with a pointer into the
38706         arc_arch_types table.
38707         (arc_selected_cpu): Declare.
38708         * config/arc/arc.c (arc_selected_cpu): Make global.
38709         (arc_selected_arch): Delete.
38710         (arc_base_cpu): Delete.
38711         (arc_override_options): Remove references to deleted variables,
38712         update access to arch information.
38713         (ARC_OPT): Update access to arch information.
38714         (ARC_OPTX): Likewise.
38715         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38716         (TARGET_ARC600): Update access to arch information.
38717         (TARGET_ARC601): Likewise.
38718         (TARGET_ARC700): Likewise.
38719         (TARGET_EM): Likewise.
38720         (TARGET_HS): Likewise.
38721         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38722         information.
38724 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38726         PR target/78604
38727         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38728         condition/operands for integer GE/LE/GEU/LEU operations.
38730 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38732         PR translation/79397
38733         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38734         of AltiVec.
38736 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38738         PR ipa/79375
38739         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38740         whether allocation happened.
38741         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38742         nothing was allocated.
38744 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38746         PR tree-optimization/79408
38747         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38748         constant, but SSA_NAME with a known integer range, use the minimum
38749         of that range instead of op1 to determine if modulo can be replaced
38750         with its first operand.
38752 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38754         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38756 2017-02-08  Richard Biener  <rguenther@suse.de>
38758         PR tree-optimization/71824
38759         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38760         Check all loops contained in the merged region.
38762 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38764         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38766 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38768         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38769         (thunderxt88): Likewise.
38770         (thunderxt81): Disable LSE and change v8.1 to v8.
38771         (thunderxt83): Likewise.
38773 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38774             Richard Biener  <rguenther@suse.de>
38776         PR middle-end/79399
38777         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38778         type from int to size_t.
38779         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38781 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38783         PR rtl-optimization/79386
38784         * cprop.c (bypass_conditional_jumps): Initialize
38785         bypass_last_basic_block already before splitting bbs after
38786         unconditional traps...
38787         (bypass_conditional_jumps): ... rather than here.
38789         PR target/79299
38790         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38791         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38792         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38793         fix -masm=intel patterns.
38795 2017-02-07  Richard Biener  <rguenther@suse.de>
38797         PR tree-optimization/79256
38798         PR middle-end/79278
38799         * builtins.c (get_object_alignment_2): Use min_align_of_type
38800         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38801         and ADJUST_FIELD_ALIGN.
38803         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38804         type parameter.
38805         * doc/tm.texi: Regenerate.
38806         * stor-layout.c (layout_decl): Adjust.
38807         (update_alignment_for_field): Likewise.
38808         (place_field): Likewise.
38809         (min_align_of_type): Likewise.
38810         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38811         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38812         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38813         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38814         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38815         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38816         * config/i386/i386.c (x86_field_alignment): Likewise.
38817         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38818         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38819         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38820         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38821         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38822         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38823          Likewise.
38825         Revert
38826         2017-01-30  Richard Biener  <rguenther@suse.de>
38828         PR tree-optimization/79256
38829         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38830         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38831         alignment on TYPE.
38833 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38835         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38836         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38837         builtins to SImode and emit a zero-extend, if necessary.
38839 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38841         * docs/invoke.texi (RISC-V Options): Alphabetize.
38843 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38845         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38846         options.
38848 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38850         * config/riscv/riscv.c: New file.
38851         * common/config/riscv/riscv-common.c: Likewise.
38852         * config.gcc: Likewise.
38853         * config/riscv/constraints.md: Likewise.
38854         * config/riscv/elf.h: Likewise.
38855         * config/riscv/generic.md: Likewise.
38856         * config/riscv/linux.h: Likewise.
38857         * config/riscv/multilib-generator: Likewise.
38858         * config/riscv/peephole.md: Likewise.
38859         * config/riscv/pic.md: Likewise.
38860         * config/riscv/predicates.md: Likewise.
38861         * config/riscv/riscv-builtins.c: Likewise.
38862         * config/riscv/riscv-c.c: Likewise.
38863         * config/riscv/riscv-ftypes.def: Likewise.
38864         * config/riscv/riscv-modes.def: Likewise.
38865         * config/riscv/riscv-opts.h: Likewise.
38866         * config/riscv/riscv-protos.h: Likewise.
38867         * config/riscv/riscv.h: Likewise.
38868         * config/riscv/riscv.md: Likewise.
38869         * config/riscv/riscv.opt: Likewise.
38870         * config/riscv/sync.md: Likewise.
38871         * config/riscv/t-elf-multilib: Likewise.
38872         * config/riscv/t-linux: Likewise.
38873         * config/riscv/t-linux-multilib: Likewise.
38874         * config/riscv/t-riscv: Likewise.
38875         * configure.ac: Likewise.
38876         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
38877         Waterman as RISC-V maintainers.
38878         * doc/install.texi: Add RISC-V entries.
38879         * doc/invoke.texi: Add RISC-V options section.
38880         * doc/md.texi: Add RISC-V constraints section.
38881         * configure: Regenerated.
38883 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
38885         PR target/66144
38886         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
38887         false values to be constant vectors with all 0 or all 1 bits set.
38888         (vcondu<mode><mode>): Likewise.
38889         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
38890         predicate.
38891         (fpmask_comparison_operator): Update comment.
38892         (vecint_comparison_operator): New predicate.
38893         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
38894         vector conditionals when the true and false values are constant
38895         vectors with all 0 bits or all 1 bits set.
38897 2017-02-06  Martin Sebor  <msebor@redhat.com>
38899         PR  tree-optimization/79376
38900         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
38902 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
38904         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
38905         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
38906         to simplify split condition.
38908 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38910         * omp-expand.c (oxpand_omp_atomic_fetch_op,
38911         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
38912         false.
38914 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
38916         PR rtl-optimization/68664
38917         * target.def (can_speculate_insn): New hook.
38918         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
38919         * doc/tm.texi: Regenerate.
38920         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
38921         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
38922         (rs6000_sched_can_speculate_insn): New function.
38924 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38926         PR tree-optimization/79284
38927         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
38928         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
38929         vectorizable_mask_load_store, vectorizable_operation,
38930         vect_is_simple_cond, get_same_sized_vectype): Use it instead
38931         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
38932         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
38933         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
38934         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
38935         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
38936         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
38937         is_gimple_assign (stmt).  Replace another such test with
38938         is_gimple_assign (stmt).
38940 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
38942         PR target/78883
38943         * config/avr/avr.c (rtl-iter.h): Include it.
38944         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
38945         (avr_legitimate_combined_insn): ...and implementation.
38947 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38949         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
38950         * config/s390/s390.c (s390_const_operand_ok)
38951         (s390_canonicalize_comparison, s390_extract_part)
38952         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
38953         (s390_contiguous_bitmask_p, s390_rtx_costs)
38954         (legitimize_pic_address): Likewise.
38955         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
38956         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
38957         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
38958         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
38959         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
38961 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
38963         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
38964         REGNO($0) == REGNO($1).
38966 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38968         * config/s390/linux.h(SIZE_TYPE): Add comment.
38970 2017-02-06  Julian Brown  <julian@codesourcery.com>
38971             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
38972             Virendra Pathak  <virendra.pathak@broadcom.com>
38974         * config/aarch64/aarch64-cores.def: Change the scheduler
38975         to Thunderx2t99.
38976         * config/aarch64/aarch64.md: Include thunderx2t99.md.
38977         * config/aarch64/thunderx2t99.md: New file.
38979 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
38981         * doc/standards.texi (Go Language): Update link to language
38982         standard.
38984 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
38986         * tree-eh.c (lower_resx): Sanitize profile.
38987         (cleanup_empty_eh_move_lp): Likewise.
38989 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
38991         PR tree-ssa/79347
38992         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
38993         ELSE_PROB.
38994         * cfgloopmanip.h (loop_version): Update prototype.
38995         * modulo-sched.c (sms_schedule): Update call of loop_version.
38996         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
38997         * tree-parloops.c (gen_parallel_loop): Likewise.
38998         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
38999         * tree-ssa-loop-split.c (split_loop): Likewise.
39000         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39001         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39003 2017-02-05  Martin Liska  <mliska@suse.cz>
39005         PR bootstrap/78985
39006         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39007         variable to NULL.
39008         (print_operand_address): Initialize a struct to zero.
39010 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39012         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39013         garbage collector only in textual form.
39015 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39017         * doc/extend.texi (x86 specific memory model extensions for
39018         transactional memory): Simplify a phrase.
39020 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39022         PR target/79353
39023         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39024         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39025         (atomic_storedi_1): Likewise.
39027 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39029         PR tree-optimization/79338
39030         * tree-parloops.c (gather_scalar_reductions): Don't call
39031         vect_analyze_loop_form for loop->inner before destroying loop's
39032         loop_vinfo.
39034 2017-02-03  Martin Sebor  <msebor@redhat.com>
39036         PR tree-optimization/79327
39037         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39038         when precision has resulted in leading zeros.
39039         (format_integer): Adjust the likely counter to assume an unknown
39040         argument that may be zero is non-zero.
39042 2017-02-03  Jason Merrill  <jason@redhat.com>
39044         PR c++/78689
39045         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39046         avoid copying non-taken branch.
39048 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39050         PR tree-optimization/79340
39051         * tree-vect-loop.c (vectorizable_reduction): Release
39052         vec_defs elements after safe_splicing them into other vectors.
39053         Formatting fixes.
39055         PR tree-optimization/79327
39056         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39057         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39058         dirtype.
39059         (format_integer): Use wide_int_to_tree instead of build_int_cst
39060         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39061         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39062         of shortest and longest sequence.
39064 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39066         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39067         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39069 2017-02-03  Walter Lee  <walt@tilera.com>
39071         PR target/78862
39072         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39073         after initial stackframe link reg save.
39074         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39076 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39078         PR target/79354
39079         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39080         wu for stxssp alternative.
39082 2017-02-03  Martin Sebor  <msebor@redhat.com>
39084         PR tree-optimization/79352
39085         * gimple-fold.c (get_range_strlen): Add argument.
39086         (get_range_strlen): Change return type to bool.
39087         (get_maxval_strlen): Pass in a dummy argument.
39088         * gimple-fold.h (get_range_strlen): Change return type to bool.
39089         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39090         * tree.h (array_at_struct_end_p): Add argument.
39091         * tree.c (array_at_struct_end_p): Handle it.
39093 2017-02-03  Martin Liska  <mliska@suse.cz>
39095         PR lto/66295
39096         * multiple_target.c (create_dispatcher_calls): Redirect edge
39097         from a caller of a dispatcher.
39098         (expand_target_clones): Make the clones local.
39099         (ipa_target_clone): Do both target clones and resolvers.
39100         (ipa_dispatcher_calls): Remove the pass.
39101         (pass_dispatcher_calls::gate): Likewise.
39102         (make_pass_dispatcher_calls): Likewise.
39103         * passes.def (pass_target_clone): Put as very first IPA early
39104         pass.
39106 2017-02-03  Martin Liska  <mliska@suse.cz>
39108         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39109         in case of a function with ifunc attribute.
39111 2017-02-03  Martin Liska  <mliska@suse.cz>
39113         * cgraph.c (cgraph_node::dump): Dump function version info.
39114         * symtab.c (symtab_node::dump_base): Add missing new line.
39116 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39118         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39119         (ifcombine_ifandif): Use it.
39121 2017-02-03  Martin Liska  <mliska@suse.cz>
39123         * doc/invoke.texi: Document default value for
39124         use-after-scope-direct-emission-threshold.
39126 2017-02-03  Martin Liska  <mliska@suse.cz>
39128         PR tree-optimization/79339
39129         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39130         (format_floating): Likewise.
39132 2017-02-03  Martin Liska  <mliska@suse.cz>
39134         PR ipa/79337
39135         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39136         implementation.
39137         (ipa_node_params_t::remove): Likewise.
39138         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39139         initialization from removed ipa_node_params_t::insert.
39140         (ipa_node_params::~ipa_node_params): Move from removed
39141         ipa_node_params_t::release.
39142         * symbol-summary.h (symbol_summary::m_released): New member.
39143         Do not release a summary twice.  Do not allow to call finalizer
39144         for types of a summary that live in GGC memory.
39146 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39148         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39149         cmp_branch fusion.
39151 2017-02-02  Martin Sebor  <msebor@redhat.com>
39153         PR middle-end/79275
39154         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39155         (format_string): Tighten up the range of output for non-constant
39156         strings and correct the expected range for wide non-constant strings.
39158 2017-02-02  Martin Sebor  <msebor@redhat.com>
39160         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39162         PR middle-end/32003
39163         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39164         index.
39165         (-fdump-tree-@var): Add to index and document how to come up
39166         with pass-specific option and dump file names.
39167         (-fdump-passes): Clarify where to look for output.
39169 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39171         PR middle-end/77445
39172         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39173         statistics of the analyzed path; allow threading for speed when
39174         any of BBs along the path are optimized for speed.
39176 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39178         PR middle-end/78468
39179         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39180         settings of the virtual registers.
39182         Revert again
39183         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39185         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39186         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39187         needed.
39189 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39191         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39192         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39194 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39196         * config/s390/s390.md: Add missing comments with the expanded
39197         mnemonics.
39198         * config/s390/vector.md: Likewise.
39199         * config/s390/vx-builtins.md: Likewise.
39201 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39203         PR target/79197
39204         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39205         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39206         conditions on a single line.
39208 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39210         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39211         __S390_VX__ to __VX__.
39213 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39215         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39216         stmt_info to record_stmt_cost.
39217         (vect_get_known_peeling_cost): Pass stmt_info if known to
39218         record_stmt_cost.
39219         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39220         cpu_vector_cost field into
39221         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39222         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39223         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39224         splitting of scalar_stmt_cost and vec_stmt_cost.
39225         (thunderx_vector_cost): Likewise.
39226         (cortexa57_vector_cost): LIkewise.
39227         (exynosm1_vector_cost): Likewise.
39228         (xgene1_vector_cost): Likewise.
39229         (thunderx2t99_vector_cost): Improve after the splitting of the two
39230         fields.
39231         (aarch64_builtin_vectorization_cost): Update for the splitting of
39232         scalar_stmt_cost and vec_stmt_cost.
39234 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39235             Richard Henderson  <rth@redhat.com>
39237         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39238         conditional on existance of a fast atomic load.
39239         * optabs-query.c (can_atomic_load_p): New function.
39240         * optabs-query.h (can_atomic_load_p): Declare it.
39241         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39242         no fast atomic load is available for the particular size of access.
39243         (expand_atomic_compare_and_swap): Likewise.
39244         (expand_atomic_load): Likewise.
39245         (expand_atomic_store): Likewise.
39246         (expand_atomic_fetch_op): Likewise.
39247         * testsuite/lib/target-supports.exp
39248         (check_effective_target_sync_int_128): Remove x86 because it provides
39249         no fast atomic load.
39250         (check_effective_target_sync_int_128_runtime): Likewise.
39252 2017-02-01  Richard Biener  <rguenther@suse.de>
39254         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39255         (graphite_transform_loops): Provide opt-info for optimized nests.
39256         * tree-parloop.c (parallelize_loops): Provide opt-info for
39257         parallelized loops.
39259 2017-02-01  Richard Biener  <rguenther@suse.de>
39261         PR middle-end/79315
39262         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39263         was not set before.
39265 2017-02-01  Richard Biener  <rguenther@suse.de>
39267         PR tree-optimization/71824
39268         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39269         Verify the loops are valid in the merged SESE region.
39270         (scop_detection::can_represent_loop_1): Check analyzing the
39271         evolution of the number of iterations in the region succeeds.
39273 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39275         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39276         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39278 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39280         PR preprocessor/79210
39281         * input.c (get_substring_ranges_for_loc): Replace line_width
39282         assertion with error-handling.
39284 2017-01-31  Richard Biener  <rguenther@suse.de>
39286         PR tree-optimization/77318
39287         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39288         (create_pw_aff_from_tree): Take loop parameter.
39289         (add_condition_to_pbb): Pass loop of the condition to
39290         create_pw_aff_from_tree.
39292 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39294         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39295         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39297 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39299         PR target/78597
39300         PR target/79038
39301         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39302         no longer used.
39303         (convert_int_to_float128): Likewise.
39304         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39305         (convert_int_to_float128): Likewise.
39306         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39307         (UNSPEC_IEEE128_CONVERT): Likewise.
39308         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39309         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39310         Use local variables for IBM extended format.
39311         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39312         (fix_trunc<mode>si2_fprs): Likewise.
39313         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39314         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39315         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39316         to know that we can now have integers of all sizes in vector
39317         registers.
39318         (fix<uns>_<mode>di2_hw): Likewise.
39319         (float<uns>_<mode>si2_hw): Likewise.
39320         (fix_<mode>si2_hw): Likewise.
39321         (fixuns_<mode>si2_hw): Likewise.
39322         (float<uns>_<mode>di2_hw): Likewise.
39323         (float_<mode>di2_hw): Likewise.
39324         (float_<mode>si2_hw): Likewise.
39325         (floatuns_<mode>di2_hw): Likewise.
39326         (floatuns_<mode>si2_hw): Likewise.
39327         (xscvqp<su>wz_<mode>): Delete, no longer used.
39328         (xscvqp<su>dz_<mode>): Likewise.
39329         (xscv<su>dqp_<mode>): Likewise.
39330         (ieee128_mfvsrd_64bit): Likewise.
39331         (ieee128_mfvsrd_32bit): Likewise.
39332         (ieee128_mfvsrwz): Likewise.
39333         (ieee128_mtvsrw): Likewise.
39334         (ieee128_mtvsrd_64bit): Likewise.
39335         (ieee128_mtvsrd_32bit): Likewise.
39337 2017-01-31  Martin Liska  <mliska@suse.cz>
39339         PR ipa/79285
39340         * ipa-prop.c (ipa_free_all_node_params): Call release method
39341         instead of ~sumbol_summary to not to trigger double times
39342         dtor of hash_map.
39344 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39346         PR tree-optimization/71691
39347         * bitmap.h (class auto_bitmap): New.
39348         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39349         is_maybe_undefined instead of ssa_undefined_value_p.
39351 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39353         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39354         __S390_ARCH_LEVEL__ to __ARCH__.
39356 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39358         PR tree-optimization/79267
39359         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39360         if should_remove_lhs_p is true.
39362 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39364         PR debug/63238
39365         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39366         (add_alignment_attribute): New.
39367         (base_type_die): Add alignment attribute.
39368         (subrange_type_die): Likewise.
39369         (modified_type_die): Likewise.
39370         (gen_array_type_die): Likewise.
39371         (gen_descr_array_type_die: Likewise.
39372         (gen_enumeration_type_die): Likewise.
39373         (gen_subprogram_die): Likewise.
39374         (gen_variable_die): Likewise.
39375         (gen_field_die): Likewise.
39376         (gen_ptr_to_mbr_type_die): Likewise.
39377         (gen_struct_or_union_type_die): Likewise.
39378         (gen_subroutine_type_die): Likewise.
39379         (gen_typedef_die): Likewise.
39380         (base_type_cmp): Compare alignment attribute.
39382 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39384         PR target/79170
39385         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39386         (setb_unsigned) New pattern for setb with CCUNS.
39387         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39388         subfc./subfe sequence to avoid overflow problems.  Generate a
39389         shorter sequence with cmpld/setb for power9.
39390         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39391         for generating subfc. instruction.
39392         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39393         now uses this instruction.
39395 2017-01-30  Ian Lance Taylor  <iant@google.com>
39397         PR debug/79289
39398         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39399         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39401 2017-01-30  Martin Sebor  <msebor@redhat.com>
39403         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39404         Move constant to the right of a relational operator.
39405         (get_mpfr_format_length, format_character, format_string): Ditto.
39406         (should_warn_p, maybe_warn): Same.
39408         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39410 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39412         PR lto/79061
39413         * asan.c (get_translation_unit_decl): Remove function.
39414         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39416 2017-01-30  Martin Liska  <mliska@suse.cz>
39418         PR gcov-profile/79259
39419         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39420         -fprofile-generate.
39422 2017-01-30  Martin Liska  <mliska@suse.cz>
39424         PR bootstrap/78985
39425         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39426         Initialize variables with NULL value.
39428 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39430         PR target/79260
39431         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39432         tm_p_file.
39433         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39435 2017-01-30  Richard Biener  <rguenther@suse.de>
39437         PR tree-optimization/79276
39438         * tree-vrp.c (process_assert_insertions): Properly adjust common
39439         when removing a duplicate.
39441 2017-01-30  Richard Biener  <rguenther@suse.de>
39443         PR tree-optimization/79256
39444         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39445         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39446         alignment on TYPE.
39447         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39449 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39451         PR target/79240
39452         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39453         ("*r<noxa>sbg_<mode>_sll_bitmask")
39454         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39455         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39456         Use contiguous_bitmask_nowrap_operand.
39458 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39460         PR target/79268
39461         * config/rs6000/altivec.h (vec_xl): Revise #define.
39462         (vec_xst): Likewise.
39464 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39466         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39468 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39470         PR rtl-optimization/79194
39471         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39472         traps before call to bypass_conditional_jumps.
39474 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39476         PR tree-optimization/71374
39477         * lra-constraints.c (check_conflict_input_operands): New.
39478         (match_reload): Use it.
39480 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39482         PR target/79131
39483         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39484         account to calculate conflict_set.
39486 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39488         PR rtl-optimization/78559
39489         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39490         other_insn in combine.
39492 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39494         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39495         uint16_type_node for BT_UINT16.
39497 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39499         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39500         "RTL Tests" to menu.
39501         (GIMPLE Tests): New node.
39502         (RTL Tests): New node.
39504 2017-01-27  Richard Biener  <rguenther@suse.de>
39506         PR tree-optimization/79245
39507         * tree-loop-distribution.c (distribute_loop): Apply cost
39508         modeling also to detected patterns.
39510 2017-01-27  Richard Biener  <rguenther@suse.de>
39512         PR tree-optimization/71433
39513         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39514         (compare_assert_loc): New function.
39515         (process_assert_insertions): Sort and optimize assert locations
39516         to remove duplicates and push down identical assertions on
39517         edges to their destination block.
39519 2017-01-27  Richard Biener  <rguenther@suse.de>
39521         PR tree-optimization/79244
39522         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39523         out SSA names even if abnormal.
39525 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39527         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39528         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39529         instead of MPFR_RNDN.
39531 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39533         PR target/79239
39534         * arm.c (arm_option_override): Don't call build_target_option_node
39535         until after doing all option overrides.
39536         (arm_valid_target_attribute_tree): Likewise.
39538 2017-01-27  Martin Liska  <mliska@suse.cz>
39540         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39541         for {cd}tors and C++ {cd}tors.
39543 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39545         * config/s390/s390.md ("*setmem_long_and")
39546         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39548 2017-01-26  Martin Sebor  <msebor@redhat.com>
39550         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39551         of precision.
39553 2017-01-26  Martin Sebor  <msebor@redhat.com>
39555         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39556         HAVE_DFmode before using XFmode or DFmode.
39557         (parse_directive): Avoid using the z length modifier to avoid
39558         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39560         PR middle-end/78703
39561         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39562         to accept adjustment as an array.
39563         (get_int_range): New function.
39564         (struct directive): Make width and prec arrays.
39565         (directive::set_width, directive::set_precision): Call get_int_range.
39566         (format_integer, format_floating): Handle width and precision ranges.
39567         (format_string, parse_directive): Same.
39569 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39571         PR debug/79129
39572         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39573         comdat_type_p set, just clone them, but keep the children in the
39574         original DIE.
39576         PR debug/78835
39577         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39578         which have direct callers with -fvar-tracking-assignments enabled
39579         in the current TU.
39580         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39581         inside of type units.
39583 2017-01-26  Martin Sebor  <msebor@redhat.com>
39585         PR middle-end/78703
39586         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39587         unlikely counters.
39588         (struct format_result): Replace number_chars, number_chars_min,
39589         and number_chars_max with a single member of struct result_range.
39590         Remove bounded.
39591         (format_result::operator+=): Adjust.
39592         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39593         counters.
39594         (fmtresult::adjust_for_width_or_precision): New function.
39595         (fmtresult:type_max_digits): New function.
39596         (bytes_remaining): Handle likely and unlikely counters.
39597         (min_bytes_remaining): Remove.
39598         (format_percent): Simplify.
39599         (format_integer, format_floating): Set likely and unlikely counters.
39600         (get_string_length, format_character, format_string): Same.
39601         (format_plain, should_warn_p): New function.
39602         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39603         and handle those for all directives, including plain strings.
39604         (format_directive): Handle likely and unlikely counters.
39605         Remove unnecessary quoting from diagnostics.  Add an informational
39606         note.
39607         (add_bytes): Remove.
39608         (pass_sprintf_length::compute_format_length): Simplify.
39609         (try_substitute_return_value): Handle likely and unlikely counters.
39611 2017-01-26  Carl Love  <cel@us.ibm.com>
39613         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39614         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39616 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39618         PR target/79131
39619         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39620         endianess for subregs into account.
39621         * lra-constraints.c (lra_constraints): Do risky transformations
39622         always on the first iteration.
39623         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39624         last_call_used_reg_set.
39625         (process_bb_lives): Define and use last_call_used_reg_set.
39626         * lra.c (lra): Always continue after lra_constraints on the first
39627         iteration.
39629 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39631         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39632         constant.
39633         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39635 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39637         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39638         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39639         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39640         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39641         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39642         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39643         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39644         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39645         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39647 2017-01-26  Marek Polacek  <polacek@redhat.com>
39649         PR c/79199
39650         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39651         for the third operand.
39653 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39655         PR middle-end/79236
39656         * omp-low.c (struct omp_context): Add simt_stmt field.
39657         (scan_omp_for): Return omp_context *.
39658         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39659         context to the _simt_ SIMD stmt.
39660         (lower_omp_for): For combined SIMD with sibling _simt_
39661         SIMD, make sure to use the same decls in _looptemp_
39662         clauses as in the sibling.
39664 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39666         PR middle-end/79212
39667         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39668         all contexts.
39670 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39672         PR target/70465
39673         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39674         emit fld b; fld a; if possible.
39676         * brig-builtins.def: Update copyright years.
39677         * config/arm/arm_acle_builtins.def: Update copyright years.
39679 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39681         PR target/79179
39682         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39683         constraint instead of o for the stxsd instruction.
39685 2017-01-25  Carl Love  <cel@us.ibm.com>
39687         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39688         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39690 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39692         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39694 2017-01-25  Richard Biener  <rguenther@suse.de>
39696         PR tree-optimization/69264
39697         * target.def (vector_alignment_reachable): Improve documentation.
39698         * doc/tm.texi: Regenerate.
39699         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39700         and add a comment.
39701         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39702         earlier changes with respect to TYPE_USER_ALIGN.
39703         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39705 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39707         PR target/79145
39708         * config/arm/arm.md (xordi3): Force constant operand into a register
39709         for TARGET_IWMMXT.
39711 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39713         * doc/invoke.texi (-fstore-merging): Correct default optimization
39714         levels at which it is enabled.
39715         (-O): Move -fstore-merging from list to...
39716         (-O2): ... Here.
39718 2017-01-25  Richard Biener  <rguenther@suse.de>
39720         PR debug/78363
39721         * omp-expand.c: Include debug.h.
39722         (expand_omp_taskreg): Make sure to generate early debug before
39723         outlining anything from a function.
39724         (expand_omp_target): Likewise.
39725         (grid_expand_target_grid_body): Likewise.
39727 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39729         PR lto/79061
39730         * asan.c (get_translation_unit_decl): New function.
39731         (asan_add_global): Extract modules file name from globals
39732         TRANSLATION_UNIT_DECL name.
39734 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39736         PR target/77439
39737         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39738         for long calls with APCS frame and VFP.
39740 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39742         * cfg.c (original_copy_tables_initialized_p): New function.
39743         * cfg.h (original_copy_tables_initialized_p): New decl.
39744         * cfgrtl.c (relink_block_chain): Guard the call to
39745         free_original_copy_tables with a call to
39746         original_copy_tables_initialized_p.
39747         * cgraph.h (symtab_node::native_rtl_p): New decl.
39748         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39749         (symtab_node::needed_p): Don't assert for early assembly output
39750         for __RTL functions.
39751         (cgraph_node::finalize_function): Set "force_output" for __RTL
39752         functions.
39753         (cgraph_node::analyze): Bail out early for __RTL functions.
39754         (analyze_functions): Update assertion to support __RTL functions.
39755         (cgraph_node::expand): Bail out early for __RTL functions.
39756         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39757         __RTL functions.
39758         * function.h (struct function): Update comment for field
39759         "pass_startwith".
39760         * gimple-expr.c: Include "tree-pass.h".
39761         (gimple_has_body_p): Return false for __RTL functions.
39762         * Makefile.in (OBJS): Add run-rtl-passes.o.
39763         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39764         accessor.
39765         (gcc::pass_manager::get_clean_slate): New accessor.
39766         * passes.c: Include "insn-addr.h".
39767         (should_skip_pass_p): Add logging.  Update logic for running
39768         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39769         property-provider override so it is only done for gimple passes.
39770         Don't skip dfinit.
39771         (skip_pass): New function.
39772         (execute_one_pass): Call skip_pass when skipping passes.
39773         * read-md.c (md_reader::read_char): Support filtering
39774         the input to a subset of line numbers.
39775         (md_reader::md_reader): Initialize fields
39776         m_first_line and m_last_line.
39777         (md_reader::read_file_fragment): New function.
39778         * read-md.h (md_reader::read_file_fragment): New decl.
39779         (md_reader::m_first_line): New field.
39780         (md_reader::m_last_line): New field.
39781         * read-rtl-function.c (function_reader::create_function): Only
39782         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39783         curr_properties.  Set DECL_INITIAL to a dummy block.
39784         (read_rtl_function_body_from_file_range): New function.
39785         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39786         New decl.
39787         * run-rtl-passes.c: New file.
39788         * run-rtl-passes.h: New file.
39790 2017-01-24  Jeff Law  <law@redhat.com>
39792         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39793         buffer size.
39795 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39797         PR tree-optimization/79159
39798         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39799         (record_nonwrapping_iv): Improve boundary using above function if no
39800         value range information.
39802 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39803             Martin Jambor  <mjambor@suse.cz>
39805         * brig-builtins.def: New file.
39806         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39807         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39808         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39809         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39810         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39811         * builtin-types.def (BT_INT8): New.
39812         (BT_INT16): Likewise.
39813         (BT_UINT8): Likewise.
39814         (BT_UINT16): Likewise.
39815         (BT_FN_ULONG): Likewise.
39816         (BT_FN_UINT_INT): Likewise.
39817         (BT_FN_UINT_ULONG): Likewise.
39818         (BT_FN_UINT_LONG): Likewise.
39819         (BT_FN_UINT_PTR): Likewise.
39820         (BT_FN_ULONG_PTR): Likewise.
39821         (BT_FN_INT8_FLOAT): Likewise.
39822         (BT_FN_INT16_FLOAT): Likewise.
39823         (BT_FN_UINT32_FLOAT): Likewise.
39824         (BT_FN_UINT16_FLOAT): Likewise.
39825         (BT_FN_UINT8_FLOAT): Likewise.
39826         (BT_FN_UINT64_FLOAT): Likewise.
39827         (BT_FN_UINT16_UINT32): Likewise.
39828         (BT_FN_UINT32_UINT16): Likewise.
39829         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39830         (BT_FN_INT_PTR_INT): Likewise.
39831         (BT_FN_UINT_PTR_UINT): Likewise.
39832         (BT_FN_LONG_PTR_LONG): Likewise.
39833         (BT_FN_ULONG_PTR_ULONG): Likewise.
39834         (BT_FN_VOID_UINT64_UINT64): Likewise.
39835         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39836         (BT_FN_INT8_INT8_INT8): Likewise.
39837         (BT_FN_INT16_INT16_INT16): Likewise.
39838         (BT_FN_INT_INT_INT): Likewise.
39839         (BT_FN_UINT_FLOAT_UINT): Likewise.
39840         (BT_FN_FLOAT_UINT_UINT): Likewise.
39841         (BT_FN_ULONG_UINT_UINT): Likewise.
39842         (BT_FN_ULONG_UINT_PTR): Likewise.
39843         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39844         (BT_FN_UINT_UINT_UINT): Likewise.
39845         (BT_FN_VOID_UINT_PTR): Likewise.
39846         (BT_FN_UINT_UINT_PTR: Likewise.
39847         (BT_FN_UINT32_UINT64_PTR): Likewise.
39848         (BT_FN_INT_INT_UINT_UINT): Likewise.
39849         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39850         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39851         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39852         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39853         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39854         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39855         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39856         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39857         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39858         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39859         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39860         * doc/frontends.texi: List BRIG FE.
39861         * doc/install.texi (Testing): Add BRIG tesring requirements.
39862         * doc/invoke.texi (Overall Options): Mention BRIG.
39863         * doc/standards.texi (Standards): Doucment BRIG HSA version.
39865 2017-01-24  Richard Biener  <rguenther@suse.de>
39867         PR translation/79208
39868         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
39870 2017-01-24  Martin Jambor  <mjambor@suse.cz>
39872         PR bootstrap/79198
39873         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
39874         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
39875         and known_contexts.
39877 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
39879         PR middle-end/79123
39880         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
39881         casts from signed to unsigned really don't have a range.
39883 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
39885         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
39886         GMP_RNDx for compatiblity.
39888 2017-01-24  Martin Liska  <mliska@suse.cz>
39890         PR bootstrap/79132
39891         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
39892         that would prevent us to call alloca with -1 as argument.
39894 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
39896         * dwarf2out.c (output_compilation_unit_header, output_file_names):
39897         Avoid -Wformat-security warning.
39899 2017-01-23  Andrew Pinski  <apinski@cavium.com>
39901         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
39902         cost table.
39904 2017-01-23  Martin Sebor  <msebor@redhat.com>
39906         PR middle-end/78703
39907         * gimple-ssa-sprintf.c (warn_level): New global.
39908         (format_integer): Use it here and throughout the rest of the file.
39909         Use the same switch to compute sign as base.
39910         (maybe_warn): New function.
39911         (format_directive): Factor out warnings into maybe_warn.
39912         Add debugging output.  Use warn_level.
39913         (add_bytes): Use warn_level.
39914         (pass_sprintf_length::compute_format_length): Add debugging output.
39915         (try_substitute_return_value): Same.
39916         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
39918         PR middle-end/78703
39919         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
39920         (struct fmtresult, format_integer, format_floating): Adjust.
39921         (fmtresult::fmtresult): Set max correctly in two argument ctor.
39922         (get_string_length, format_string,format_directive): Same.
39923         (pass_sprintf_length::compute_format_length): Same.
39924         (try_substitute_return_value): Simplify slightly.
39926         PR middle-end/78703
39927         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
39928         (fmtresult::operator+=): Outlined.
39929         (struct fmtresult): Add ctors.
39930         (struct conversion_spec): Rename...
39931         (struct directive): ...to this.  Add and remove data members.
39932         (directive::set_width, directive::set_precision): New functions.
39933         (format_percent): Use fmtresult ctor.
39934         (get_width_and_precision): Remove.
39935         (format_integer): Make naming changes.  Avoid computing width and
39936         precision.
39937         (format_floating): Same.  Adjust indentation.
39938         (format_character, format_none): New functions.
39939         (format_string): Moved character handling to format_character.
39940         (format_directive): Remove arguments, change return type.
39941         (parse_directive): New function.
39942         (pass_sprintf_length::compute_format_length): Move directive
39943         parsing to parse_directive.
39945 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
39947         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
39948         (assign_assembler_name_if_needed): ... this.
39949         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
39950         (assign_assembler_name_if_needed): ... this.
39951         (free_lang_data_in_cgraph): Adjust callers.
39952         * cgraphunit.c (cgraph_node::analyze): Likewise.
39953         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
39954         Likewise.
39956 2017-01-23  Richard Biener  <rguenther@suse.de>
39958         PR tree-optimization/79088
39959         PR tree-optimization/79188
39960         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
39961         resetting loop bounds after last path deletion.  Reset loop
39962         bounds of the target loop, make code match the comments.
39963         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
39964         Make sure loops need no fixups.
39966 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
39968         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
39969         exponent support with double type for first argument.
39970         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
39971         type returned by __builtin_vec_extract_sig,
39972         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
39973         functions from "vector int" to "vector unsigned int" or from
39974         "vector long long int" to "vector unsigned long long int".
39975         Changed type returned by __builtin_vec_extract_exp,
39976         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
39977         functions from "vector int" to "vector unsigned int" or from
39978         "vector long long int" to "vector unsigned long long int".
39979         Changed return type of __builtin_vec_test_data_class,
39980         __builtin_vec_test_data_class_sp, and
39981         __builtin_vec_test_data_class_dp from "vector int" to
39982         "vector bool int" or from "vector long long int" to "vector bool
39983         long long int" and changed second argument type from "unsigned
39984         int" to "int".  Added new overloaded function forms "vector float
39985         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
39986         "vector float __builtin_vec_insert_exp_sp (vector float, vector
39987         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
39988         double, vector unsigned long long int)" and "vector double
39989         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
39990         long int)".  Changed return type of
39991         __builtin_scalar_test_data_class and
39992         __builtin_scalar_test_data_class_sp and
39993         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
39994         int" and changed second argument from "unsigned int" to "int".
39995         Changed type returned by __builtin_scalar_test_neg,
39996         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
39997         from "int" to "bool int".  Added new overloaded function form
39998         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
39999         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40000         exponent double-precision with floating point first argument.
40001         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40002         documentation of scalar_test_data_class, scalar_test_neg,
40003         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40004         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40005         vec_test_data_class built-in functions to reflect refinements in
40006         their type signatures.
40008 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40010         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40011         size of buf.
40012         (aarch64_elf_asm_destructor): Likewise.
40014 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40016         PR rtl-optimization/78634
40017         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40018         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40019         * ifcvt.c (noce_try_cmove): Add missing cost check.
40021         PR rtl-optimization/71724
40022         * combine.c (if_then_else_cond): Look for situations where it is
40023         beneficial to undo the work of one of the recursive calls.
40025 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40027         PR tree-optimization/70754
40028         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40029         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40030         combined stmt before it if not NULL.
40031         (combine_chains): Process refs reversely and compute dominance point
40032         for root ref.
40034 2017-01-23  Martin Liska  <mliska@suse.cz>
40036         PR tree-optimization/79196
40037         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40038         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40039         instead of memcmp.
40040         (strlen_optimize_stmt): Call the renamed function.
40042 2017-01-23  Michael Matz  <matz@suse.de>
40044         PR tree-optimization/78384
40045         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40047 2017-01-23  Richard Biener  <rguenther@suse.de>
40049         PR tree-optimization/79186
40050         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40051         both incoming edges before moving an assert.
40053 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40055         * ipa-prop.c (load_from_param_1): Removed.
40056         (load_from_unmodified_param): Bits from load_from_param_1 put back
40057         here.
40058         (load_from_param): Removed.
40059         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40060         with stmt.  Reverted back to use of load_from_unmodified_param.
40062 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40064         PR ipa/79108
40065         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40066         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40067         field a pointer to garbage collected vector, mark lattices and
40068         ipcp_orig_node with GTY((skip)).
40069         (ipa_get_param_count): Adjust to descriptors being a pointer.
40070         (ipa_get_param): Likewise.
40071         (ipa_get_type): Likewise.
40072         (ipa_get_param_move_cost): Likewise.
40073         (ipa_set_param_used): Likewise.
40074         (ipa_get_controlled_uses): Likewise.
40075         (ipa_set_controlled_uses): Likewise.
40076         (ipa_is_param_used): Likewise.
40077         (ipa_node_params_t): Move into garbage collector.  New methods insert
40078         and remove.
40079         (ipa_node_params_sum): Annotate wth GTY(()).
40080         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40081         garbage collected.
40082         (ipa_load_from_parm_agg): Adjust declaration.
40083         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40084         * ipa-profile.c (ipa_profile): Likewise.
40085         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40086         (ipa_populate_param_decls): Make descriptors parameter garbage
40087         collected.
40088         (ipa_dump_param): Adjust to descriptors being a pointer.
40089         (ipa_alloc_node_params): Likewise.
40090         (ipa_initialize_node_params): Likewise.
40091         (load_from_param_1): Make descriptors parameter garbage collected.
40092         (load_from_unmodified_param): Likewise.
40093         (load_from_param): Likewise.
40094         (ipa_load_from_parm_agg): Likewise.
40095         (ipa_node_params::~ipa_node_params): Removed.
40096         (ipa_free_all_node_params): Remove call to delete operator.
40097         (ipa_node_params_t::insert): New.
40098         (ipa_node_params_t::remove): Likewise.
40099         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40100         copy known_csts and known_contexts vectors.
40101         (ipa_read_node_info): Adjust to descriptors being a pointer.
40102         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40103         collected.
40104         (ipcp_transform_function): Make descriptors variable garbage
40105         collected.
40107 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40109         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40110         * config/i386/avx512dqintrin.h: Ditto.
40111         * config/i386/avx512fintrin.h: Ditto.
40112         * config/i386/i386.c: Handle new builtins.
40113         * config/i386/i386-builtin.def: Add new builtins.
40114         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40115         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40117 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40118             Martin Liska  <mliska@suse.cz>
40120         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40121         * asan.c (asan_expand_poison_ifn): Support stores and use
40122         appropriate ASAN report function.
40123         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40124         * internal-fn.def (ASAN_POISON_USE): Declare.
40125         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40126         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40127         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40128         ASAN_POISON calls w/o LHS.
40129         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40130         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40131         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40132         * gimplify.c (asan_poison_variables): Add attribute
40133         use_after_scope_memory to variables that really needs to live
40134         in memory.
40135         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40136         having the attribute.
40138 2017-01-23  Martin Liska  <mliska@suse.cz>
40140         * asan.c (create_asan_shadow_var): New function.
40141         (asan_expand_poison_ifn): Likewise.
40142         * asan.h (asan_expand_poison_ifn): New declaration.
40143         * internal-fn.c (expand_ASAN_POISON): Likewise.
40144         * internal-fn.def (ASAN_POISON): New builtin.
40145         * sanopt.c (pass_sanopt::execute): Expand
40146         asan_expand_poison_ifn.
40147         * tree-inline.c (copy_decl_for_dup_finish): Make function
40148         external.
40149         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40150         * tree-ssa.c (is_asan_mark_p): New function.
40151         (execute_update_addresses_taken): Rewrite local variables
40152         (identified just by use-after-scope as addressable) into SSA.
40154 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40156         * doc/install.texi (Specific): opensource.apple.com uses https
40157         now. Remove trailing slash.
40159 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40161         * README.Portability: Remove note on an Irix compatibility issue.
40163 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40165         * gcov.c (INCLUDE_ALGORITHM): Define.
40166         (INCLUDE_VECTOR): Define.
40167         No longer include <vector> and <algorithm> directly.
40169 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40171         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40172         to https.
40173         * doc/invoke.texi (Code Gen Options): Ditto.
40175 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40177         PR lto/78407
40178         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40180 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40182         rtl-optimization/79125
40183         * cprop.c (local_cprop_pass): Handle cases where we make an
40184         unconditional trap.
40186 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40188         PR target/61729
40189         PR target/77850
40190         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40191         read from, for big endian.
40193 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40195         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40196         register pauth builtins for LP64 only.
40198 2017-01-20  Marek Polacek  <polacek@redhat.com>
40200         PR c/79152
40201         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40202         non-case labels.
40204 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40206         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40207         of safelen status.
40208         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40209         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40210         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40212 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40214         PR target/71270
40215         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40216         in big-endian mode when they are not a single duplicated value.
40218 2017-01-20  Richard Biener  <rguenther@suse.de>
40220         * BASE-VER: Bump to 7.0.1.
40222 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40224         * omp-low.c (omplow_simd_context): New struct.  Use it...
40225         (lower_rec_simd_input_clauses): ...here and...
40226         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40227         references to idx, lane, max_vf, is_simt.
40229 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40231         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40232         mcpu=nps400.
40234 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40236         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40237         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40238         gt-hsa-common.h.
40239         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40240         (GTFILES): Rename hsa.c to hsa-common.c.
40241         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40242         * hsa-dump.c: Likewise.
40243         * hsa-gen.c: Likewise.
40244         * hsa-regalloc.c: Likewise.
40245         * ipa-hsa.c: Likewise.
40246         * omp-expand.c: Likewise.
40247         * omp-low.c: Likewise.
40248         * toplev.c: Likewise.
40250 2017-01-20  Marek Polacek  <polacek@redhat.com>
40252         PR c/64279
40253         * doc/invoke.texi: Document -Wduplicated-branches.
40254         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40255         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40256         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40257         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40258         return 0 only when not OEP_LEXICOGRAPHIC.
40259         (fold_build_cleanup_point_expr): Use the expression
40260         location when building CLEANUP_POINT_EXPR.
40261         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40262         * tree.c (add_expr): Handle error_mark_node.
40264 2017-01-20  Martin Liska  <mliska@suse.cz>
40266         PR lto/69188
40267         * tree-profile.c (init_ic_make_global_vars): Do not call
40268         finalize_decl.
40269         (gimple_init_gcov_profiler): Likewise.
40271 2017-01-20  Martin Liska  <mliska@suse.cz>
40273         PR ipa/71190
40274         * cgraph.h (maybe_create_reference): Remove argument and
40275         update comment.
40276         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40277         argument.
40278         * ipa-cp.c (create_specialized_node): Likewise.
40279         * symtab.c (symtab_node::maybe_create_reference): Handle
40280         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40282 2017-01-20  Martin Liska  <mliska@suse.cz>
40284         * read-rtl-function.c (function_reader::create_function): Use
40285         build_decl instread of build_decl_stat.
40287 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40289         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40290         * config/i386/avx512dqintrin.h: Ditto.
40291         * config/i386/avx512fintrin.h: Ditto.
40292         * config/i386/i386-builtin-types.def: Add new types.
40293         * config/i386/i386.c: Handle new types.
40294         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40295         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40296         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40297         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40298         (__builtin_ia32_kshiftridi): New.
40299         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40301 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40303         PR target/78875
40304         PR target/79140
40305         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40306         define to rs6000_init_stack_protect_guard.
40307         (rs6000_init_stack_protect_guard): New function.
40309 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40310             Yunqiang Su  <yunqiang.su@imgtec.com>
40312         * config.gcc (supported_defaults): Add madd4.
40313         (with_madd4): Add validation.
40314         (all_defaults): Add madd4.
40315         * config/mips/mips.opt (mmadd4): New option.
40316         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40317         mmadd4.
40318         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40319         __mips_no_madd4.
40320         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40321         (ISA_HAS_FUSED_MADD4): Likewise.
40322         * doc/invoke.texi (-mmadd4): Document the new option.
40323         * doc/install.texi (--with-madd4): Document the new option.
40325 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40327         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40328         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40329         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40330         (aarch64_init_pauth_hint_builtins): New.
40331         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40332         (aarch64_expand_builtin): Expand new builtins.
40334 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40336         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40337         * combine-stack-adj.c (no_unhandled_cfa): Handle
40338         REG_CFA_TOGGLE_RA_MANGLE.
40339         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40340         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40341         info for return address signing.
40342         (aarch64_expand_epilogue): Likewise.
40344 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40346         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40347         * config/aarch64/aarch64-protos.h
40348         (aarch64_return_address_signing_enabled): New declaration.
40349         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40350         New function.
40351         (aarch64_expand_prologue): Sign return address before it's pushed onto
40352         stack.
40353         (aarch64_expand_epilogue): Authenticate return address fetched from
40354         stack.
40355         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40356         (aarch64_attributes): New function attributes for "sign-return-address".
40357         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40358         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40359         ("*do_return"): Generate combined instructions according to key index.
40360         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40361         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40362         iterators.
40363         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40364         * config/aarch64/aarch64.opt (msign-return-address=): New.
40365         * doc/extend.texi (AArch64 Function Attributes): Documents
40366         "sign-return-address=".
40367         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40369 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40371         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40372         overall option summary.
40374 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40376         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40377         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40378         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40379         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40381 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40383         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40384         -mpower9-minmax by default for -mcpu=power9.
40385         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40386         128-bit floating point.
40388 2017-01-20  Alan Modra  <amodra@gmail.com>
40390         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40391         optimizing for size.
40393 2017-01-20  Alan Modra  <amodra@gmail.com>
40395         PR target/79144
40396         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40397         for strcmp and strncmp from corresponding builtin decl.
40399 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40401         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40402         instead of i386/rtems-64.h.
40403         * config/i386/rtems-64.h: Remove.
40405 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40407         PR target/78478
40408         Revert:
40409         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40411         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40413 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40415         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40416         Change int to HOST_WIDE_INT.
40417         * config/aarch64/aarch64-protos.h
40418         (aarch64_simd_gen_const_vector_dup): Likewise.
40419         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40421 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40423         * langhooks-def.h (lhd_type_for_size): New decl.
40424         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40425         * langhooks.c (lhd_type_for_size): New function, taken from
40426         lto_type_for_size.
40428 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40430         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40431         define_bypass for CR latency.
40432         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40433         (power9-alu2): Add define_bypass for CR latency.
40434         (power9-cmp): New.
40435         (power9-mul): Update insn latency.
40436         (power9-mul-compare): Update insn latency, bypass latency and remove
40437         power9-branch.
40439 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40441         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40442         Delete.
40443         * config/aarch64/aarch64.md
40444         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40445         aarch64_nopcrelative_literal_loads.
40446         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40448 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40450         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40451         TARGET_LOONGSON_3A.
40452         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40454 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40456         PR target/78176
40457         * config.gcc (supported_defaults): Add lxc1-sxc1.
40458         (with_lxc1_sxc1): Add validation.
40459         (all_defaults): Add lxc1-sxc1.
40460         * config/mips/mips.opt (mlxc1-sxc1): New option.
40461         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40462         mlxc1-sxc1.
40463         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40464         __mips_no_lxc1_sxc1.
40465         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40466         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40467         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40469 2017-01-19  Richard Biener  <rguenther@suse.de>
40471         PR tree-optimization/72488
40472         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40473         sure to restore SSA info.
40474         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40476 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40478         PR rtl-optimization/79121
40479         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40480         of the inner type when shifting an extended value.
40482 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40484         PR lto/78407
40485         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40486         interposable aliases.
40488 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40490         PR target/78516
40491         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40492         Use the evmergelohi instruction.
40493         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40494         (mov_sitf_e500_subreg8_2_be): Likewise.
40495         (mov_sitf_e500_subreg12_2_le): Likewise.
40496         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40497         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40498         (mov_sitf_e500_subreg8_2_le): Likewise.
40499         (mov_sitf_e500_subreg12_2_be): Likewise.
40501 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40503         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40504         attribute from vecsimple to vecperm.
40505         (altivec_vbpermq2): Likewise.
40507 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40509         PR target/79040
40510         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40512 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40513         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40514         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40515         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40516         case where N arg is SIZE_MAX.
40517         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40518         (cmpstrsi): Add pattern.
40520 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40522         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40523         __builtin_vec_revb builtins.
40524         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40525         built-in functions to support generation of the ISA 3.0 XXBR<x>
40526         vector byte reverse instructions.
40527         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40528         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40529         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40530         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40531         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40532         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40533         (P9V_BUILTIN_VEC_REVB): Likewise.
40534         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40535         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40536         (p9_xxbrq_v16qi): Likewise.
40537         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40538         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40539         (p9_xxbrh_v8hi): Likewise.
40540         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40541         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40542         vec_revb built-in functions.
40544 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40546         PR rtl-optimization/78952
40547         * config/i386/i386.md (any_extract): New code iterator.
40548         (*insvqi_2): Use any_extract for source operand.
40549         (*insvqi_3): Use any_shiftrt for source operand.
40551 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40553         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40554         New function.
40555         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40557 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40559         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40561 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40563         * config/rs6000/altivec.h (vec_bperm): Change #define.
40564         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40565         (altivec_vbpermq2): New define_insn.
40566         (altivec_vbpermd): Likewise.
40567         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40568         function interface.
40569         (VBPERMD): Likewise.
40570         (VBPERM): New polymorphic function interface.
40571         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40572         Add entries for P9V_BUILTIN_VEC_VBPERM.
40573         * doc/extend.texi: Add interfaces for vec_bperm.
40575 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40577         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40578         first letter of error messages.
40579         (s390_resolve_overloaded_builtin): Likewise.
40580         * config/s390/s390.c (s390_expand_builtin): Likewise.
40581         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40582         (s390_valid_target_attribute_inner_p): Likewise.
40583         * config/s390/s390.md ("tabort"): Likewise.
40585 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40587         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40588         (ISA_AVOID_DIV_HILO): New macro.
40589         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40590         (ISA_HAS_DDIV): Likewise.
40592 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40594         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40596 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40598         * doc/invoke.texi (fabi-version): Spelling fix.
40600 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40602         PR c++/70182
40603         * doc/invoke.texi (fabi-version): Mention mangling fix for
40604         operator names.
40606 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40608         PR c++/77489
40609         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40611 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40613         PR target/78875
40614         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40615         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40616         the new options.
40617         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40618         flexible settings.
40619         (stack_protect_test): Ditto.
40620         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40621         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40622         options.
40623         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40624         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40625         -mstack-protector-guard-offset=.
40626         (RS/6000 and PowerPC Options): Ditto.
40628 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40630         * config/i386/i386.h (MASK_CLASS_P): New define.
40631         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40632         there are no registers from different register sets also when
40633         mask registers are used.  Update function comment.
40634         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40635         to (*k/*r) and (*k/*km) alternatives.
40637 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40639         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40640         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40641         (EH_RETURN_HANDLER_RTX): New define.
40642         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40643         Force frame pointer in EH return functions.
40644         (aarch64_expand_epilogue): Add barrier for eh_return.
40645         (aarch64_final_eh_return_addr): Remove.
40646         (aarch64_eh_return_handler_rtx): New function.
40647         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40648         Remove.
40649         (aarch64_eh_return_handler_rtx): New prototype.
40651 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40653         * config/rs6000/altivec.h (vec_rlmi): New #define.
40654         (vec_vrlnm): Likewise.
40655         (vec_rlnm): Likewise.
40656         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40657         (UNSPEC_VRLNM): Likewise.
40658         (VIlong): New mode iterator.
40659         (altivec_vrl<VI_char>mi): New define_insn.
40660         (altivec_vrl<VI_char>nm): Likewise.
40661         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40662         function entry.
40663         (VRLDNM): Likewise.
40664         (RLNM): New polymorphic function entry.
40665         (VRLWMI): New monomorphic function entry.
40666         (VRLDMI): Likewise.
40667         (RLMI): New polymorphic function entry.
40668         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40669         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40670         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40671         vec_vrlnm.
40673 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40675         PR debug/78839
40676         * dwarf2out.c (field_byte_offset): Restore the
40677         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40678         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40679         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40680         of build2 + fold.
40682 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40684         PR ada/67205
40685         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40687 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40689         PR debug/71669
40690         * dwarf2out.c (add_data_member_location_attribute): For constant
40691         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40692         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40693         DW_AT_byte_size attributes.
40695 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40697         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40698         after forcing to constant memory when the code model is medium.
40700 2017-01-17  Julia Koval  <julia.koval@intel.com>
40702         PR target/76731
40703         * config/i386/avx512fintrin.h
40704         (_mm512_i32gather_ps): Change __addr type to void const*.
40705         (_mm512_mask_i32gather_ps): Ditto.
40706         (_mm512_i32gather_pd): Ditto.
40707         (_mm512_mask_i32gather_pd): Ditto.
40708         (_mm512_i64gather_ps): Ditto.
40709         (_mm512_mask_i64gather_ps): Ditto.
40710         (_mm512_i64gather_pd): Ditto.
40711         (_mm512_mask_i64gather_pd): Ditto.
40712         (_mm512_i32gather_epi32): Ditto.
40713         (_mm512_mask_i32gather_epi32): Ditto.
40714         (_mm512_i32gather_epi64): Ditto.
40715         (_mm512_mask_i32gather_epi64): Ditto.
40716         (_mm512_i64gather_epi32): Ditto.
40717         (_mm512_mask_i64gather_epi32): Ditto.
40718         (_mm512_i64gather_epi64): Ditto.
40719         (_mm512_mask_i64gather_epi64): Ditto.
40720         (_mm512_i32scatter_ps): Change __addr type to void*.
40721         (_mm512_mask_i32scatter_ps): Ditto.
40722         (_mm512_i32scatter_pd): Ditto.
40723         (_mm512_mask_i32scatter_pd): Ditto.
40724         (_mm512_i64scatter_ps): Ditto.
40725         (_mm512_mask_i64scatter_ps): Ditto.
40726         (_mm512_i64scatter_pd): Ditto.
40727         (_mm512_mask_i64scatter_pd): Ditto.
40728         (_mm512_i32scatter_epi32): Ditto.
40729         (_mm512_mask_i32scatter_epi32): Ditto.
40730         (_mm512_i32scatter_epi64): Ditto.
40731         (_mm512_mask_i32scatter_epi64): Ditto.
40732         (_mm512_i64scatter_epi32): Ditto.
40733         (_mm512_mask_i64scatter_epi32): Ditto.
40734         (_mm512_i64scatter_epi64): Ditto.
40735         (_mm512_mask_i64scatter_epi64): Ditto.
40736         * config/i386/avx512pfintrin.h
40737         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40738         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40739         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40740         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40741         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40742         (_mm512_prefetch_i32scatter_ps): Ditto.
40743         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40744         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40745         (_mm512_prefetch_i64scatter_pd): Ditto.
40746         (_mm512_prefetch_i64scatter_ps): Ditto.
40747         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40748         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40749         * config/i386/avx512vlintrin.h
40750         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40751         (_mm_mmask_i32gather_ps): Ditto.
40752         (_mm256_mmask_i32gather_pd): Ditto.
40753         (_mm_mmask_i32gather_pd): Ditto.
40754         (_mm256_mmask_i64gather_ps): Ditto.
40755         (_mm_mmask_i64gather_ps): Ditto.
40756         (_mm256_mmask_i64gather_pd): Ditto.
40757         (_mm_mmask_i64gather_pd): Ditto.
40758         (_mm256_mmask_i32gather_epi32): Ditto.
40759         (_mm_mmask_i32gather_epi32): Ditto.
40760         (_mm256_mmask_i32gather_epi64): Ditto.
40761         (_mm_mmask_i32gather_epi64): Ditto.
40762         (_mm256_mmask_i64gather_epi32): Ditto.
40763         (_mm_mmask_i64gather_epi32): Ditto.
40764         (_mm256_mmask_i64gather_epi64): Ditto.
40765         (_mm_mmask_i64gather_epi64): Ditto.
40766         (_mm256_i32scatter_ps): Change __addr type to void*.
40767         (_mm256_mask_i32scatter_ps): Ditto.
40768         (_mm_i32scatter_ps): Ditto.
40769         (_mm_mask_i32scatter_ps): Ditto.
40770         (_mm256_i32scatter_pd): Ditto.
40771         (_mm256_mask_i32scatter_pd): Ditto.
40772         (_mm_i32scatter_pd): Ditto.
40773         (_mm_mask_i32scatter_pd): Ditto.
40774         (_mm256_i64scatter_ps): Ditto.
40775         (_mm256_mask_i64scatter_ps): Ditto.
40776         (_mm_i64scatter_ps): Ditto.
40777         (_mm_mask_i64scatter_ps): Ditto.
40778         (_mm256_i64scatter_pd): Ditto.
40779         (_mm256_mask_i64scatter_pd): Ditto.
40780         (_mm_i64scatter_pd): Ditto.
40781         (_mm_mask_i64scatter_pd): Ditto.
40782         (_mm256_i32scatter_epi32): Ditto.
40783         (_mm256_mask_i32scatter_epi32): Ditto.
40784         (_mm_i32scatter_epi32): Ditto.
40785         (_mm_mask_i32scatter_epi32): Ditto.
40786         (_mm256_i32scatter_epi64): Ditto.
40787         (_mm256_mask_i32scatter_epi64): Ditto.
40788         (_mm_i32scatter_epi64): Ditto.
40789         (_mm_mask_i32scatter_epi64): Ditto.
40790         (_mm256_i64scatter_epi32): Ditto.
40791         (_mm256_mask_i64scatter_epi32): Ditto.
40792         (_mm_i64scatter_epi32): Ditto.
40793         (_mm_mask_i64scatter_epi32): Ditto.
40794         (_mm256_i64scatter_epi64): Ditto.
40795         (_mm256_mask_i64scatter_epi64): Ditto.
40796         (_mm_i64scatter_epi64): Ditto.
40797         (_mm_mask_i64scatter_epi64): Ditto.
40798         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40799         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40800         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40801         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40802         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40803         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40804         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40805         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40806         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40807         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40808         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40809         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40810         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40811         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40812         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40813         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40814         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40815         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40816         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40817         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40818         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40819         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40820         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40821         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40822         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40823         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40824         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40825         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40826         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40827         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40828         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40829         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40830         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40831         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40832         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40833         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40834         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40835         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40836         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40837         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40838         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40839         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40840         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40841         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40842         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40843         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40844         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40845         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40846         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40847         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40848         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40849         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40850         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40851         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40852         definitions accordingly.
40854 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
40855             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
40857         PR target/79079
40858         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40859         gen_lowpart.
40861 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
40863         PR target/79058
40864         * ira-conflicts.c (ira_build_conflicts): Update total conflict
40865         hard regs for inner regno.
40867 2017-01-17  Martin Liska  <mliska@suse.cz>
40869         PR ipa/71207
40870         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
40871         assumption and add comment.
40873 2017-01-17  Nathan Sidwell  <nathan@acm.org>
40875         * ipa-visibility.c (localize_node): New function, broken out of ...
40876         (function_and_variable_visibility): ... here. Call it.
40878 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40880         PR middle-end/77445
40881         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
40882         correctly set frequency of oudgoing edge.
40883         (duplicate_thread_path): Fix profile updating.
40885 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40887         PR other/79046
40888         * configure.ac: Add GCC_BASE_VER.
40889         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
40890         version from BASE-VER file.
40891         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
40892         (gcc.o): Depend on $(BASEVER).
40893         * common.opt (dumpfullversion): New option.
40894         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
40895         * doc/invoke.texi: Document -dumpfullversion.
40896         * doc/install.texi: Document --with-gcc-major-version-only.
40897         * configure: Regenerated.
40899 2017-01-17  Richard Biener  <rguenther@suse.de>
40901         PR tree-optimization/71433
40902         * tree-vrp.c (register_new_assert_for): Merge same asserts
40903         on all incoming edges.
40904         (process_assert_insertions_for): Handle insertions at the
40905         beginning of BBs.
40907 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
40909         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
40910         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
40912 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
40914         PR target/78633
40915         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
40916         RTL sharing.
40918 2017-01-17  Alan Modra  <amodra@gmail.com>
40920         PR target/79066
40921         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
40922         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
40923         symbolic stack limit when pic.
40925 2017-01-16  Martin Sebor  <msebor@redhat.com>
40927         PR tree-optimization/78608
40928         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
40930 2017-01-16  Jeff Law  <law@redhat.com>
40932         Revert:
40933         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
40934         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
40935         for several include directories that may be relative to sysroot.
40936         * config/i386/x-mingw32 (gplus_includedir): Define.
40937         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
40938         (native_system_includedir): Likewise.
40939         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
40940         override if TARGET_SYSTEM_ROOT is defined.
40941         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
40943         PR tree-optimization/79090
40944         PR tree-optimization/33562
40945         PR tree-optimization/61912
40946         PR tree-optimization/77485
40947         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
40948         and computed trims into the dump file.
40950 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40952         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
40954 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
40956         PR c/79089
40957         * gimplify.c (gimplify_init_constructor): If want_value and
40958         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
40959         fix.
40961         PR target/79080
40962         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
40963         sequence.  Formatting fixes.
40964         (doloop_optimize): Formatting fixes.
40966         PR driver/49726
40967         * gcc.c (debug_level_greater_than_spec_func): New function.
40968         (static_spec_functions): Add debug-level-gt spec function.
40969         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
40970         !g0.
40971         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40972         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40973         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
40974         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
40975         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
40976         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
40978 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
40980         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
40981         QImode fixups to general and mask registers only.
40983 2017-01-16  Carl Love  <cel@us.ibm.com>
40985         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
40986         for built-in functions
40987         vector signed char vec_nabs (vector signed char)
40988         vector signed short vec_nabs (vector signed short)
40989         vector signed int vec_nabs (vector signed int)
40990         vector signed long long vec_nabs (vector signed long long)
40991         vector float vec_nabs (vector float)
40992         vector double vec_nabs (vector double)
40993         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
40994         and NABS overload.
40995         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
40996         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
40997         * doc/extend.texi: Update the documentation file for the new built-in
40998         functions.
41000 2017-01-16  Martin Sebor  <msebor@redhat.com>
41002         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41003         message.
41005 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41007         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41008         UNSPEC_VSX__XXSPLTD to require special splat handling.
41010 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41012         PR bootstrap/78616
41013         * system.h: Poison strndup.
41015 2017-01-16  Alan Modra  <amodra@gmail.com>
41017         PR target/79098
41018         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41019         use a switch.
41021 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41023         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41025 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41027         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41028         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41030 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41032         PR target/72749
41033         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41034         fallthrough.
41035         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41036         in the currently scheduled RTL fragment.
41038 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41040         PR rtl-optimization/78751
41041         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41042         give up.
41044 2017-01-14  Jeff Law  <law@redhat.com>
41046         PR tree-optimization/79090
41047         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41048         variable length stores.
41049         (compute_trims): Delete dead assignment to *trim_tail.
41050         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41051         zero length.
41053 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41055         PR rtl-optimization/78626
41056         PR rtl-optimization/78727
41057         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41058         of a block, and split such blocks after everything else is finished.
41060 2017-01-14  Alan Modra  <amodra@gmail.com>
41062         PR target/72749
41063         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41064         target legitimate_combined_insn.
41065         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41066         (rs6000_legitimate_combined_insn): New function.
41067         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41068         all uses.
41069         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41070         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41071         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41073 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41075         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41077 2017-01-13  Jeff Law  <law@redhat.com>
41079         PR tree-optimization/33562
41080         PR tree-optimization/61912
41081         PR tree-optimization/77485
41082         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41083         a statement.
41084         (delete_dead_assignment): Likewise.
41085         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41086         statement to delete_dead_call and delete_dead_assignment.
41088 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41090         PR c/78304
41091         * substring-locations.c (format_warning_va): Strengthen case 1 so
41092         that both endpoints of the substring must be within the format
41093         range for just the substring to be printed.
41095 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41097         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41098         * config/i386/i386.c (ix86_target_string): Add missing options
41099         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41100         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41101         flags_other and ix86_target_other to flags2_other.  Display unknown
41102         isa2 options.
41103         (ix86_valid_target_attribute_inner_p): Add missing options and
41104         reorder options by implied ISAs, as in ix86_target_string.
41106 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41108         * hash-table.h (hash_table::too_empty_p): New function.
41109         (hash_table::expand): Use it.
41110         (hash_table::traverse): Likewise.
41111         (hash_table::empty_slot): Use sizeof (value_type) instead of
41112         sizeof (PTR) to convert bytes to elements.  Shrink the table
41113         if the current size is excessive for the current number of
41114         elements.
41116 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41118         * ira-costs.c (record_reg_classes): Break from the inner loop
41119         early once alt_fail is known to be true.  Update outer loop
41120         handling accordingly.
41122 2017-01-13  Jeff Law  <law@redhat.com>
41124         * tree-ssa-dse.c (decrement_count): New function.
41125         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41126         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41127         when we know the partially dead statement is a mem* function.
41129         PR tree-optimization/61912
41130         PR tree-optimization/77485
41131         * tree-ssa-dse.c: Include expr.h.
41132         (maybe_trim_constructor_store): New function.
41133         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41135         PR tree-optimization/33562
41136         PR tree-optimization/61912
41137         PR tree-optimization/77485
41138         * doc/invoke.texi: Document new dse-max-object-size param.
41139         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41140         * tree-ssa-dse.c: Include params.h.
41141         (dse_store_status): New enum.
41142         (initialize_ao_ref_for_dse): New, partially extracted from
41143         dse_optimize_stmt.
41144         (valid_ao_ref_for_dse, normalize_ref): New.
41145         (setup_live_bytes_from_ref, compute_trims): Likewise.
41146         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41147         (maybe_trim_partially_dead_store): Likewise.
41148         (maybe_trim_complex_store): Likewise.
41149         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41150         Track what bytes live from the original store.  Return tri-state
41151         for dead, partially dead or live.
41152         (dse_dom_walker): Add constructor, destructor and new private members.
41153         (delete_dead_call, delete_dead_assignment): New extracted from
41154         dse_optimize_stmt.
41155         (dse_optimize_stmt): Make a member of dse_dom_walker.
41156         Use initialize_ao_ref_for_dse.
41158         PR tree-optimization/33562
41159         PR tree-optimization/61912
41160         PR tree-optimization/77485
41161         * sbitmap.h (bitmap_count_bits): Prototype.
41162         (bitmap_clear_range, bitmap_set_range): Likewise.
41163         * sbitmap.c (bitmap_clear_range): New function.
41164         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41166 2017-01-13  Martin Liska  <mliska@suse.cz>
41168         PR ipa/79043
41169         * function.c (set_cfun): Add new argument force.
41170         * function.h (set_cfun): Likewise.
41171         * ipa-inline-transform.c (inline_call): Use the function when
41172         strict alising from is dropped for function we inline to.
41174 2017-01-13  Richard Biener  <rguenther@suse.de>
41176         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41177         for dumping GIMPLE INTEGER_CSTs.
41179 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41181         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41182         to 201112L since C++17.
41184 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41186         PR sanitizer/78887
41187         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41188         if -fsanitize=kernel-address is present.
41190 2017-01-13  Richard Biener  <rguenther@suse.de>
41192         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41193         as _Literal ( type ) number in case usual suffixes do not
41194         preserve all information.
41196 2017-01-13  Richard Biener  <rguenther@suse.de>
41198         PR tree-optimization/77283
41199         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41200         and ssa-iterators.h.
41201         (is_feasible_trace): Implement a cost model based on joiner
41202         PHI node uses.
41204 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41206         PR target/79004
41207         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41208         char or short to __float128/_Float128 directly.
41210 2017-01-12  Martin Sebor  <msebor@redhat.com>
41212         to -Wformat-overflow.
41213         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41214         (min_bytes_remaining): Same.
41215         (get_string_length): Same.
41216         (format_string): Same.
41217         (format_directive): Same.
41218         (add_bytes): Same.
41219         (pass_sprintf_length::handle_gimple_call): Same.
41221 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41223         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41224         info.nowrite calls with no lhs that can't throw.  Return bool
41225         whether gsi_remove has been called or not.
41226         (pass_sprintf_length::handle_gimple_call): Return bool whether
41227         try_substitute_return_value called gsi_remove.  Formatting fix.
41228         (pass_sprintf_length::execute): Don't use gsi_remove if
41229         handle_gimple_call returned true.
41231         PR bootstrap/79069
41232         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41233         be removed due to side-effects, don't remove following barrier nor
41234         turn the successor edge into fallthru edge.
41236 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41238         PR target/79044
41239         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41240         element-reversing loads and stores as not swappable.
41242 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41243             Nicolai Stange  <nicstange@gmail.com>
41245         * combine.c (try_combine): Don't ignore result of overlap checking
41246         loop.  Combine overlap & asm check into single loop.
41248 2017-01-12  Richard Biener  <rguenther@suse.de>
41250         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41251         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41253 2017-01-12  Richard Biener  <rguenther@suse.de>
41255         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41256         and TS_TARGET_OPTION directly derive from TS_BASE.
41257         * tree-core.h (tree_optimization_option): Derive from tree_base.
41258         (tree_target_option): Likewise.
41260 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41262         * config/i386/i386.c (memory_address_length): Increase len
41263         only when rip_relative_addr_p returns false.
41265 2017-01-11  Julia Koval  <julia.koval@intel.com>
41267         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41268         (OPTION_MASK_ISA_SGX_SET): New.
41269         (ix86_handle_option): Handle OPT_msgx.
41270         * config.gcc: Added sgxintrin.h.
41271         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41272         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41273         * config/i386/i386.c (ix86_target_string): Add -msgx.
41274         (PTA_SGX): New.
41275         (ix86_option_override_internal): Handle new options.
41276         (ix86_valid_target_attribute_inner_p): Add sgx.
41277         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41278         * config/i386/i386.opt: Add msgx.
41279         * config/i386/sgxintrin.h: New file.
41280         * config/i386/x86intrin.h: Add sgxintrin.h.
41282 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41284         PR c++/71537
41285         * fold-const.c (maybe_nonzero_address): Return 1 for function
41286         local objects.
41287         (tree_single_nonzero_warnv_p): Don't handle function local objects
41288         here.
41290         PR c++/72813
41291         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41292         of c-header.
41294 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41296         PR driver/78877
41297         * opts.c: Include "spellcheck.h"
41298         (struct string_fragment): New struct.
41299         (struct edit_distance_traits<const string_fragment &>): New
41300         struct.
41301         (get_closest_sanitizer_option): New function.
41302         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41304 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41306         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41307         by 12.
41308         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41309         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41310         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41311         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41312         for initial die_offset if dwarf_split_debug_info.
41313         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41314         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41315         fields.
41316         (output_skeleton_debug_sections): Formatting fix.  Use
41317         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41318         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41320 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41322         * config/arm/cortex-a53.md: Add bypasses for
41323         cortex_a53_r2f_cvt.
41324         (cortex_a53_r2f): Only use for transfers.
41325         (cortex_a53_f2r): Likewise.
41326         (cortex_a53_r2f_cvt): Add reservation for conversions.
41327         (cortex_a53_f2r_cvt): Likewise.
41329 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41331         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41332         to all inlined functions, change static to extern.
41334 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41336         PR target/78253
41337         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41338         weak symbol.
41339         (arm_assemble_integer): Likewise.
41341 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41343         * config.gcc: Use new awk script to check CPU, FPU and architecture
41344         parameters for --with-... options.
41345         * config/arm/parsecpu.awk: New file
41346         * config/arm/arm-cpus.in: New file.
41347         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41348         files.
41349         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41350         files.
41351         * config/arm/t-arm: Update dependency rules.
41352         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41353         of processing .def files.
41354         * config/arm/genopt.sh: Deleted.
41355         * config/arm/gentune.sh: Deleted.
41356         * config/arm/arm-cores.def: Deleted.
41357         * config/arm/arm-arches.def: Deleted.
41358         * config/arm/arm-fpus.def: Deleted.
41359         * config/arm/arm-tune.md: Regenerated.
41360         * config/arm/arm-tables.opt: Regenerated.
41361         * config/arm/arm-cpu.h: New generated file.
41362         * config/arm/arm-cpu-data.h: New generated file.
41363         * config/arm/arm-cpu-cdata.h: New generated file.
41365 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41367         PR lto/79042
41368         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41369         bit.
41370         (input_varpool_node): Unpack dynamically_initialized bit.
41372 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41374         PR rtl-optimization/79032
41375         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41376         the alignment of the adjusted memory reference against that of MODE,
41377         instead of the alignment of the original memory reference.
41379 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41381         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41382         test.
41383         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41384         decorated functions.
41386 2017-01-11  Richard Biener  <rguenther@suse.de>
41388         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41389         set range/nonnull info for PHI results.  Do not set it on
41390         stmts marked for removal.
41392 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41394         * expr.c (store_field): In the bitfield case, fetch the return value
41395         from the registers before applying a single big-endian adjustment.
41396         Always do a final load for a BLKmode value not larger than a word.
41398 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41400         PR c++/77949
41401         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41402         that we correctly handle column numbers greater than
41403         LINE_MAP_MAX_COLUMN_NUMBER.
41405 2017-01-10  Martin Sebor  <msebor@redhat.com>
41407         PR middle-end/78245
41408         * gimple-ssa-sprintf.c (get_destination_size): Call
41409         {init,fini}object_sizes.
41410         * tree-object-size.c (addr_object_size): Adjust.
41411         (pass_through_call): Adjust.
41412         (pass_object_sizes::execute): Adjust.
41413         * tree-object-size.h (fini_object_sizes): Declare.
41415 2017-01-10  Martin Sebor  <msebor@redhat.com>
41417         PR tree-optimization/78775
41418         * builtins.c (get_size_range): Move...
41419         * calls.c: ...to here.
41420         (alloc_max_size): Accept zero argument.
41421         (operand_signed_p): Remove.
41422         (maybe_warn_alloc_args_overflow): Call get_size_range.
41423         * calls.h (get_size_range): Declare.
41425 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41427         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41428         from TI's devices.csv file as of September 2016.
41429         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41431 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41433         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41434         * doc/invoke.texi: Likewise.
41435         * doc/md.texi: Likewise.
41436         * doc/objc.texi: Likewise.
41438 2017-01-10  Joshua Conner  <joshconner@google.com>
41440         * config/arm/fuchsia-elf.h: New file.
41441         * config/fuchsia.h: New file.
41442         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41443         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41444         targets.
41445         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41447 2016-01-10  Richard Biener  <rguenther@suse.de>
41449         PR tree-optimization/79034
41450         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41451         Propagate out degenerate PHIs in the joiner.
41453 2017-01-10  Martin Liska  <mliska@suse.cz>
41455         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41456         (sort_congruence_classes_by_decl_uid): Likewise.
41457         (sort_congruence_class_groups_by_decl_uid): Likewise.
41458         (sem_item_optimizer::merge_classes): Sort class, groups in these
41459         classes and members in the groups by DECL_UID of declarations.
41460         This would make merge operations stable.
41462 2017-01-10  Martin Liska  <mliska@suse.cz>
41464         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41465         usage of m_classes_vec.
41466         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41467         (sem_item_optimizer::get_group_by_hash): Likewise.
41468         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41469         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41470         (sem_item_optimizer::verify_classes): Likewise.
41471         (sem_item_optimizer::process_cong_reduction): Likewise.
41472         (sem_item_optimizer::dump_cong_classes): Likewise.
41473         (sem_item_optimizer::merge_classes): Likewise.
41474         * ipa-icf.h (congruence_class_hash): Rename from
41475         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41477 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41479         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41480         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41481         * config.gcc: Add avx512vpopcntdqintrin.h.
41482         * config/i386/avx512vpopcntdqintrin.h: New.
41483         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41484         * config/i386/i386-builtin-types.def: Add new types.
41485         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41486         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41487         __builtin_ia32_vpopcountq_v8di_mask): New.
41488         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41489         __AVX512VPOPCNTDQ__.
41490         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41491         (PTA_AVX512VPOPCNTDQ): Define.
41492         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41493         TARGET_AVX512VPOPCNTDQ_P): Define.
41494         * config/i386/i386.opt: Add mavx512vpopcntdq.
41495         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41496         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41498 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41500         PR middle-end/77484
41501         * predict.def (PRED_CALL): Set to 67.
41503 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41505         * expr.c (store_field): In the bitfield case, if the value comes from
41506         a function call and is of an aggregate type returned in registers, do
41507         not modify the field mode; extract the value in all cases if the mode
41508         is BLKmode and the size is not larger than a word.
41510 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41512         PR target/71017
41513         * config/i386/cpuid.h: Fix undefined behavior.
41515 2017-01-04  Jeff Law  <law@redhat.com>
41517         PR tree-optimization/79007
41518         PR tree-optimization/67955
41519         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41520         conservative for pt.null when flag_non_call_exceptions is on.
41522 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41524         PR translation/79019
41525         PR translation/79020
41526         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41527         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41528         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41529         in descriptions.
41530         * config/avr/avr.opt (maccumulate-args): Likewise.
41531         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41532         * common.opt (freport-bug): Likewise.
41533         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41534         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41535         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41536         translatable string.
41537         * config/i386/i386.c (function_value_32): Likewise.
41538         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41539         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41540         Likewise.
41541         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41542         * common/config/msp430/msp430-common.c (msp430_handle_option):
41543         Likewise.
41544         * symtab.c (symtab_node::verify_base): Likewise.
41545         * opts.c (set_debug_level): Likewise.
41546         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41547         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41548         missing whitespace to translatable strings.
41549         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41550         * config/sh/superh.h: Likewise.
41551         * config/i386/xopintrin.h: Likewise.
41552         * config/i386/znver1.md: Likewise.
41553         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41554         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41555         * double-int.h (struct double_int): Likewise.
41556         * double-int.c (div_and_round_double): Likewise.
41557         * wide-int.cc: Likewise.
41558         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41559         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41560         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41561         (crossjumps_occurred): ... this.
41562         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41563         Adjust all uses.
41565         PR tree-optimization/78899
41566         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41567         returning bool return struct loop *, NULL for failure and the new
41568         loop on success.
41569         (versionable_outer_loop_p): Don't version outer loop if it has
41570         dont_vectorized bit set.
41571         (tree_if_conversion): When versioning outer loop, ensure
41572         tree_if_conversion is performed also on the inner loop of the
41573         non-vectorizable outer loop copy.
41574         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41575         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41576         prevent vectorization of it.
41577         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41578         the outer loop vectorization of the non-scalar version is attempted
41579         before vectorization of the inner loop in scalar version.  If
41580         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41581         vectorization of its inner loop.
41582         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41583         has 2 inner loops, rename also on edges from bb whose single pred
41584         is outer_loop->header.  Fix typo in function comment.
41586 2017-01-09  Martin Sebor  <msebor@redhat.com>
41588         PR bootstrap/79033
41589         * asan.c (asan_emit_stack_protection): Increase local buffer size
41590         to avoid snprintf truncation warning.
41592 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41594         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41595         to reference thunderx2t99 for the tuning structure
41596         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41597         Rename to ...
41598         (thunderx2t99_extra_costs): This.
41599         * config/aarch64/aarch64-tune.md: Regenerate.
41600         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41601         (vulcan_addrcost_table): This.
41602         (vulcan_regmove_cost): Rename to ...
41603         (thunderx2t99_regmove_cost): This.
41604         (vulcan_vector_cost): Rename to ...
41605         (thunderx2t99_vector_cost): this.
41606         (vulcan_branch_cost): Rename to ...
41607         (thunderx2t99_branch_cost): This.
41608         (vulcan_tunings): Rename to ...
41609         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41610         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41612 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41614         PR ipa/78365
41615         PR ipa/78599
41616         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41617         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41618         (propagate_vr_accross_jump_function): Use the above function for all
41619         value range computations for pass-through jump functions and type
41620         converasion from explicit value range values.
41621         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41622         parameters from TYPE_ARG_TYPES.
41623         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41624         (ipa_write_node_info): Stream type of the actual argument.
41625         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41627 2017-01-09  Martin Liska  <mliska@suse.cz>
41629         PR pch/78970
41630         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41631         (lookup_compiler): Do not show error message with have_E.
41633 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41635         PR tree-optimization/78938
41636         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41637         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41638         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41639         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41640         fixes.
41642 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41644         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41645         is const0_rtx.
41647 2017-01-09  Richard Biener  <rguenther@suse.de>
41649         PR tree-optimization/78997
41650         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41651         name condition properly.
41653 2017-01-09  Richard Biener  <rguenther@suse.de>
41655         PR debug/79000
41656         * dwarf2out.c (is_cxx): New overload with context.
41657         (is_naming_typedef_decl): Use it.
41659 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41661         * invoke.texi (Option Summary): Correct spacing in option lists
41662         and add line breaks to fix over-long lines.
41664 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41666         PR middle-end/17660
41668         * extend.texi (Common Variable Attributes): Add xref to GCC
41669         Internals manual to explain mode attribute keywords.
41671 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41673         PR other/16519
41674         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41675         and Preprocessor Options.
41676         (Options for Linking): Document -pthread here....
41677         (RS/6000 and PowerPC Options): ...not here.
41678         (Solaris 2 Options): ...or here.
41679         * doc/cppopts.texi: Document -pthread.
41681 2017-01-08  Martin Sebor  <msebor@redhat.com>
41683         PR middle-end/77708
41684         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41685         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41686         New member functions.
41687         (format_directive): Used them.
41688         (add_bytes): Same.
41689         (pass_sprintf_length::handle_gimple_call): Same.
41690         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41691         to avoid truncation for any argument.
41692         (extract_affine_mul): Same.
41693         * tree.c (get_file_function_name): Same.
41695 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41697         PR middle-end/77484
41698         * predict.def (PRED_INDIR_CALL): Set to 86.
41700 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41702         PR preprocessor/54124
41703         * doc/cppopts.texi: Reformat -d subtable to list the full name
41704         of the options.  Add cross-reference to the docs for the general
41705         compiler -d options.
41706         * doc/invoke.texi (Developer Options): Add cross-reference to the
41707         preprocessor-specific -d option documentation.
41709 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41711         PR preprocessor/13498
41712         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41713         redudant material, and reflect new command-line options.
41714         (System Headers): Likewise.
41716 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41718         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41719         -isystem, and -idirafter.  Copy-edit.
41720         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41721         default for -ftrack-macro-expansion.  Delete obsolete and
41722         badly-formatted implementation details about -fdebug-cpp output.
41723         * doc/cppwarnopts.texi: Copy-edit.
41725 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41727         PR c++/72803
41728         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41729         that the transition from a max line width >= 1<<10 to narrower
41730         lines works correctly.
41732 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41734         * doc/options.texi (PerFunction): New.
41735         * opt-functions.awk (switch_flags): Map both Optimization and
41736         PerFunction to CL_OPTIMIZATION.
41737         * opth-gen.awk: Test for PerFunction flag along with
41738         Optimization.
41739         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41740         it only when the latter is present.  Skip those that don't in
41741         the hash function generator.
41742         * common.opt (fvar-tracking): Mark as PerFunction instead of
41743         Optimization.
41744         (fvar-tracking-assignments): Likewise.
41745         (fvar-tracking-assignments-toggle): Likewise.
41746         (fvar-tracking-uninit): Likewise.
41748 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41750         PR translation/79018
41751         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41752         the and store.
41754 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41756         PR target/57583
41757         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41758         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41759         TARGET_LONG_JUMP_TABLE_OFFSETS.
41760         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41761         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41762         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41763         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41764         * config/m68k/m68k.md (tablejump expander): Likewise.
41765         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41766         TARGET_LONG_JUMP_TABLE_OFFSETS.
41767         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41768         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41770 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41771             David Holsgrove <david.holsgrove@xilinx.com>
41773         * common/config/microblaze/microblaze-common.c
41774         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41775         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41776         New prototype.
41777         * config/microblaze/microblaze.c (microblaze_must_save_register)
41778         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41779         calls_eh_return.
41780         (microblaze_eh_return): New function.
41781         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41782         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41783         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41784         * config/microblaze/microblaze.md (eh_return): New pattern.
41786 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41788         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41789         GCC_DIAGNOSTIC_STRINGIFY): Define.
41791         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41793 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41795         * config/arm/arm.md (<mcrr>): New.
41796         (<mrrc>): New.
41797         * config/arm/arm.c (arm_arch5te): New.
41798         (arm_option_override): Set arm_arch5te.
41799         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41800         and mrrc2.
41801         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41802         (arm_mcrr_qualifiers): ... this. New.
41803         (MRRC_QUALIFIERS): Define to...
41804         (arm_mrrc_qualifiers): ... this. New.
41805         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41806         __arm_mrrc2): New.
41807         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41808         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41809         (MRRCI, mrrc, MRRC): New.
41810         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41811         VUNSPEC_MRRC2): New.
41813 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41815         * config/arm/arm.md (<mcr>): New.
41816         (<mrc>): New.
41817         * config/arm/arm.c (arm_coproc_builtin_available): Add
41818         support for mcr, mrc, mcr2 and mrc2.
41819         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41820         (arm_mcr_qualifiers): ... this. New.
41821         (MRC_QUALIFIERS): Define to ...
41822         (arm_mrc_qualifiers): ... this. New.
41823         (MCR_QUALIFIERS): Define to ...
41824         (arm_mcr_qualifiers): ... this. New.
41825         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41826         __arm_mrc2): New.
41827         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41828         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41829         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41830         VUNSPEC_MRC2): New.
41832 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41834         * config/arm/arm.md (*ldc): New.
41835         (*stc): New.
41836         (<ldc>): New.
41837         (<stc>): New.
41838         * config/arm/arm.c (arm_coproc_builtin_available): Add
41839         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41840         (arm_coproc_ldc_stc_legitimate_address): New.
41841         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41842         'qualifier_const_pointer'.
41843         (LDC_QUALIFIERS): Define to...
41844         (arm_ldc_qualifiers): ... this. New.
41845         (STC_QUALIFIERS): Define to...
41846         (arm_stc_qualifiers): ... this. New.
41847         * config/arm/arm-protos.h
41848         (arm_coproc_ldc_stc_legitimate_address): New.
41849         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41850         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41851         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41852         stc2, stcl, stc2l): New.
41853         * config/arm/constraints.md (Uz): New.
41854         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41855         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41856         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41857         VUNSPEC_STC2L): New.
41859 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41861         * config/arm/arm.md (<cdp>): New.
41862         * config/arm/arm.c (neon_const_bounds): Rename this ...
41863         (arm_const_bounds): ... this.
41864         (arm_coproc_builtin_available): New.
41865         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
41866         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
41867         (CDP_QUALIFIERS): Define to...
41868         (arm_cdp_qualifiers): ... this. New.
41869         (void_UP): Define.
41870         (arm_expand_builtin_args): Add case for 6 arguments.
41871         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
41872         (arm_const_bounds): ... this.
41873         (arm_coproc_builtin_available): New.
41874         * config/arm/arm_acle.h (__arm_cdp): New.
41875         (__arm_cdp2): New.
41876         * config/arm/arm_acle_builtins.def (cdp): New.
41877         (cdp2): New.
41878         * config/arm/iterators.md (CDPI,CDP,cdp): New.
41879         * config/arm/neon.md: Rename all 'neon_const_bounds' to
41880         'arm_const_bounds'.
41881         * config/arm/types.md (coproc): New.
41882         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
41883         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
41884         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
41885         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
41887 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41889         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
41890         (UBINOP_QUALIFIERS): New.
41891         (si_UP): Define.
41892         (acle_builtin_data): New. Change comment.
41893         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
41894         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
41895         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
41896         arm_acle_builtins.def.
41897         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
41898         (arm_init_acle_builtins): New.
41899         (CRC32_BUILTIN): Remove.
41900         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
41901         crc32cb, crc32ch and crc32cw.
41902         (arm_init_crc32_builtins): Remove.
41903         (arm_init_builtins): Use arm_init_acle_builtins rather
41904         than arm_init_crc32_builtins.
41905         (arm_expand_acle_builtin): New.
41906         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
41907         * config/arm/arm_acle_builtins.def: New.
41909 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41911         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
41912         (arm_builtin_datum): ... this.
41913         (arm_init_neon_builtin): Rename to ...
41914         (arm_init_builtin): ... this. Add a new parameters PREFIX
41915         and USE_SIG_IN_NAME.
41916         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
41917         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
41918         'arm_builtin_datum'.
41919         (arm_init_vfp_builtins): Likewise.
41920         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
41921         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
41922         (arm_expand_neon_args): Rename to ...
41923         (arm_expand_builtin_args): ... this. Rename builtin_arg
41924         enum values and differentiate between ARG_BUILTIN_MEMORY
41925         and ARG_BUILTIN_NEON_MEMORY.
41926         (arm_expand_neon_builtin_1): Rename to ...
41927         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
41928         values, arm_expand_builtin_args and add bool parameter NEON.
41929         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
41930         (arm_expand_vfp_builtin): Likewise.
41931         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
41933 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41935         PR middle-end/77484
41936         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
41937         * predict.c (tree_estimate_probability_bb): Reverse direction of
41938         polymorphic call predictor.
41940 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
41942         * passes.c (execute_one_pass): Split out pass-skipping logic into...
41943         (determine_pass_name_match): ...this new function and...
41944         (should_skip_pass_p): ...this new function.
41946 2017-01-06  Nathan Sidwell  <nathan@acm.org>
41948         * ipa-visibility.c (function_and_variable_visibility): Reformat
41949         comments and long lines.  Remove extrneous if.
41950         * symtab.c (symtab_node::make_decl_local): Fix code format.
41951         (symtab_node::set_section_for_node): Fix comment typo.
41953 2017-01-06  Martin Liska  <mliska@suse.cz>
41955         PR bootstrap/79003
41956         * lra-constraints.c: Rename invariant to lra_invariant.
41957         * predict.c (set_even_probabilities): Initialize e to NULL.
41959 2017-01-05  Martin Sebor  <msebor@redhat.com>
41961         PR tree-optimization/78910
41962         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
41963         (format_integer): Correct off-by-one error in the handling
41964         of precision with negative numbers in signed conversions..
41966 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
41968         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
41970 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
41972         PR tree-optimization/71016
41973         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
41974         factor_out_conditional_conversion.  Formatting fix.
41975         (factor_out_conditional_conversion): Add cond_stmt argument.
41976         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
41977         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
41978         Formatting fix.
41980 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
41982         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
41983         read-rtl-function.o, and selftest-rtl.o.
41984         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
41985         (selftest::aarch64_test_loading_full_dump): New function.
41986         (selftest::aarch64_run_selftests): New function.
41987         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
41988         selftest::aarch64_run_selftests.
41989         * config/i386/i386.c
41990         (selftest::ix86_test_loading_dump_fragment_1): New function.
41991         (selftest::ix86_test_loading_call_insn): New function.
41992         (selftest::ix86_test_loading_full_dump): New function.
41993         (selftest::ix86_test_loading_unspec): New function.
41994         (selftest::ix86_run_selftests): Call the new functions.
41995         * emit-rtl.c (maybe_set_max_label_num): New function.
41996         * emit-rtl.h (maybe_set_max_label_num): New decl.
41997         * function.c (instantiate_decls): Guard call to
41998         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
41999         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42000         "static".
42001         * gensupport.c (gen_reader::gen_reader): Pass "false"
42002         for new "compact" param of rtx_reader.
42003         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42004         rather than an empty string for NULL strings.
42005         * read-md.c: Potentially include config.h rather than bconfig.h.
42006         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42007         (have_error): New global, copied from errors.c.
42008         (md_reader::read_name): Rename to...
42009         (md_reader::read_name_1): ...this, adding "out_loc" param,
42010         and converting "missing name or number" to returning false, rather
42011         than failing.
42012         (md_reader::read_name): Reimplement in terms of read_name_1.
42013         (md_reader::read_name_or_nil): New function.
42014         (md_reader::read_string): Handle "(nil)" by returning NULL.
42015         (md_reader::md_reader): Add new param "compact".
42016         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42017         (md_reader::read_file): New method.
42018         * read-md.h (md_reader::md_reader): Add new param "compact".
42019         (md_reader::read_file): New method.
42020         (md_reader::is_compact): New accessor.
42021         (md_reader::read_name): Convert return type from void to file_location.
42022         (md_reader::read_name_or_nil): New decl.
42023         (md_reader::read_name_1): New decl.
42024         (md_reader::m_compact): New field.
42025         (noop_reader::noop_reader): Pass "false" for new "compact" param
42026         of rtx_reader.
42027         (rtx_reader::rtx_reader): Add new "compact" param.
42028         (rtx_reader::read_rtx_operand): Make virtual and convert return
42029         type from void to rtx.
42030         (rtx_reader::read_until): New decl.
42031         (rtx_reader::handle_any_trailing_information): New virtual function.
42032         (rtx_reader::postprocess): New virtual function.
42033         (rtx_reader::finalize_string): New virtual function.
42034         (rtx_reader::m_in_call_function_usage): New field.
42035         (rtx_reader::m_reuse_rtx_by_id): New field.
42036         * read-rtl-function.c: New file.
42037         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42038         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42039         (selftest::verify_three_block_rtl_cfg): New decl.
42040         * read-rtl-function.h: New file.
42041         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42042         For host, include function.h, memmodel.h, and emit-rtl.h.
42043         (one_time_initialization): New function.
42044         (struct compact_insn_name): New struct.
42045         (compact_insn_names): New array.
42046         (find_code): Handle insn codes in compact dumps.
42047         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42048         (bind_subst_iter_and_attr): Likewise.
42049         (add_condition_to_string): Likewise.
42050         (add_condition_to_rtx): Likewise.
42051         (apply_attribute_uses): Likewise.
42052         (add_current_iterators): Likewise.
42053         (apply_iterators): Likewise.
42054         (initialize_iterators): Guard usage of apply_subst_iterator with
42055         #ifdef GENERATOR_FILE.
42056         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42057         (md_reader::read_mapping): Likewise.
42058         (add_define_attr_for_define_subst): Likewise.
42059         (add_define_subst_attr): Likewise.
42060         (read_subst_mapping): Likewise.
42061         (check_code_iterator): Likewise.
42062         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42063         logic to...
42064         (one_time_initialization): New function.
42065         (rtx_reader::read_until): New method.
42066         (read_flags): New function.
42067         (parse_reg_note_name): New function.
42068         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42069         Handle reuse_rtx ids.
42070         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42071         Add parsing support for RTL dumps, mirroring the special-cases in
42072         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42073         values, and calling handle_any_trailing_information.
42074         (rtx_reader::read_rtx_operand): Convert return type from void
42075         to rtx, returning return_rtx.  Handle case 'e'.  Call
42076         finalize_string on XSTR and XTMPL fields.
42077         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42078          "(nil)" values were omitted.  Call the postprocess vfunc on the
42079         return_rtx.
42080         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42081         class ctor.  Initialize m_in_call_function_usage.  Call
42082         one_time_initialization.
42083         * rtl-tests.c (selftest::test_uncond_jump): Call
42084         set_new_first_and_last_insn.
42085         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42086         * selftest-rtl.c: New file.
42087         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42088         (selftest::get_insn_by_uid): New decl.
42089         * selftest-run-tests.c (selftest::run_tests): Call
42090         read_rtl_function_c_tests.
42091         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42092         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42093         dumps.
42095 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42097         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42098         operands in a special way.  Assert that pos+len <= mode precision.
42100 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42102         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42103         3 argument Alias with unlimited for the negative form.
42104         (fno-vect-cost-model): Removed.
42106 2017-01-05  Martin Liska  <mliska@suse.cz>
42108         * hsa-gen.c (gen_hsa_divmod): New function.
42109         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42111 2017-01-05  Martin Liska  <mliska@suse.cz>
42113         PR pch/78970
42114         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42115         header.
42117 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42119         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42120         small constant length operands.
42122 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42124         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42125         between loop iterations.
42127 2017-01-05  Martin Liska  <mliska@suse.cz>
42129         PR sanitizer/78815
42130         * gimplify.c (gimplify_decl_expr): Compare to
42131         asan_poisoned_variables instread of checking flags.
42132         (gimplify_target_expr): Likewise.
42133         (gimplify_expr): Likewise.
42134         (gimplify_function_tree): Conditionally initialize
42135         asan_poisoned_variables.
42137 2017-01-04  Jeff Law  <law@redhat.com>
42139         PR tree-optimizatin/78812
42140         * rtl.h (contains_mem_rtx_p): Prototype.
42141         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42142         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42143         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42144         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42145         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42147 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42149         * input.c (assert_char_at_range): Default-initialize actual_range.
42151 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42153         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42154         to match the caller.
42156 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42158         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42159         insns after final jump in test to emit dummy move.
42161 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42163         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42164         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42166 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42168         * multiple_target.c (create_dispatcher_calls): Init e_next.
42169         * tree-ssa-loop-split.c (split_loop): Init border.
42170         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42171         scalar_type.
42173 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42175         PR target/71977
42176         PR target/70568
42177         PR target/78823
42178         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42179         (altivec_register_operand): Do not return true if the operand
42180         contains a SUBREG mixing SImode and SFmode.
42181         (vsx_register_operand): Likewise.
42182         (vsx_reg_sfsubreg_ok): New predicate.
42183         (vfloat_operand): Do not return true if the operand contains a
42184         SUBREG mixing SImode and SFmode.
42185         (vint_operand): Likewise.
42186         (vlogical_operand): Likewise.
42187         (gpc_reg_operand): Likewise.
42188         (int_reg_operand): Likewise.
42189         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42190         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42191         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42192         SImode and SFmode.
42193         (rs6000_emit_move_si_sf_subreg): New helper function.
42194         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42195         fixup SUBREGs involving SImode and SFmode.
42196         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42197         numbers for the new peephole2 optimization.
42198         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42199         the GLIBC math library that do AND/IOR/XOR operations on single
42200         precision floating point.
42201         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42202         target macros to say whether we need to avoid SUBREGs mixing
42203         SImode and SFmode.
42204         (TARGET_ALLOW_SF_SUBREG): Likewise.
42205         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42206         (UNSPEC_SI_FROM_SF): Likewise.
42207         (iorxor): Change spacing.
42208         (and_ior_xor): New iterator for AND, IOR, and XOR.
42209         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42210         (movdi_from_sf_zero_ext): Likewise.
42211         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42212         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42213         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42214         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42215         (fms<mode>4): Likewise.
42216         (fnma<mode>4): Likewise.
42217         (fnms<mode>4): Likewise.
42218         (nfma<mode>4): Likewise.
42219         (nfms<mode>4): Likewise.
42221 2017-01-04  Marek Polacek  <polacek@redhat.com>
42223         PR c++/64767
42224         * doc/invoke.texi: Document -Wpointer-compare.
42226 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42228         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42229         RejectNegative.
42231         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42232         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42233         2-byte data.
42235 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42237         PR target/78056
42238         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42239         documentation of the powerpc_popcntb_ok attribute.
42240         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42241         code to issue warning messages if a requested CPU configuration is
42242         not supported by the binary (assembler and loader) toolchain.
42243         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42244         made to define a built-in function that has been disabled.
42245         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42246         made to define a built-in function that has been disabled.
42247         (altivec_init_builtins): Add comment explaining why definition
42248         of the DST built-in functions is not preceded by an assertion
42249         check.  Add assertions to prevent ICE if attempts are made to
42250         define an altivec predicate or an abs* built-in function that has
42251         been disabled.
42252         (htm_init_builtins): Add comment explaining why definition of the
42253         htm built-in functions is not preceded by an assertion check.
42255 2017-01-04  Jeff Law  <law@redhat.com>
42257         PR tree-optimizatin/67955
42258         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42259         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42260         the points-to solution does not include pt_null.  Use DECL_PT_UID
42261         unconditionally.
42263 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42265         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42266         Use gen_int_mode instead of gen_lopwart for const_int operands.
42268 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42270         PR tree-optimization/71563
42271         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42272         out of range value - has low bits known to be zero.
42274 2017-01-04  Alan Modra  <amodra@gmail.com>
42276         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42277         * configure: Regenerate.
42278         * config.in: Regenerate.
42280 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42282         PR bootstrap/77569
42283         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42284         a substring of the message, but strcmp with the whole message.  Ifdef
42285         ENABLE_NLS, translate the message first using dgettext.
42287 2017-01-03  Jeff Law  <law@redhat.com>
42289         PR tree-optimizatin/78856
42290         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42291         (mark_threaded_blocks): Remove code to truncate thread paths that
42292         cross multiple loop headers.  Instead invalidate the cached loop
42293         iteration information and handle case of a thread path walking
42294         into an irreducible region.
42296 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42298         PR target/78900
42299         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42300         assertions.  Add support for doing the signbit if the IEEE 128-bit
42301         floating point value is in a GPR.
42302         * config/rs6000/rs6000.md (Fsignbit): Delete.
42303         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42304         Update the length attribute if the value is in a GPR.
42305         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42306         the sign or zero extension instruction, since the value is always 0/1.
42307         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42309         PR target/78953
42310         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42311         extracting SImode to a GPR register so that we can generate a
42312         store, limit the vector to be in a traditional Altivec register
42313         for the vextuwrx instruction.
42315 2017-01-03  Ian Lance Taylor  <iant@google.com>
42317         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42319 2017-01-03  Martin Sebor  <msebor@redhat.com>
42321         PR tree-optimization/78696
42322         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42323         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42324         of %g.
42325         (pass_sprintf_length::compute_format_length): Set width and precision
42326         specified by asrerisk to void_node for vararg functions.
42327         (try_substitute_return_value): Adjust dump output.
42329 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42331         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42333 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42335         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42336         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42337         -mlra/-mno-lra was passed to the compiler.
42339 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42341         PR rtl-optimization/65618
42342         * emit-rtl.c (try_split): Move initialization of "before" and
42343         "after" to just before the call to emit_insn_after_setloc.
42345 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42347         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42349 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42351         * dwarf2out.c (gen_enumeration_type_die): When
42352         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42354 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42356         PR tree-optimization/78965
42357         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42358         Change first argument from const call_info & to call_info &.  For %n
42359         set info.nowrite to false.
42361         PR middle-end/78901
42362         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42363         possibly throwing calls.
42365         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42366         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42367         and fns handling, rather than in a separate case SSA_NAME.
42369 2017-01-02  Jeff Law  <law@redhat.com>
42371         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42372         fixes for first_period and second_period variables.
42374 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42376         PR target/78967
42377         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42378         (*insvqi_1): New insn pattern.
42379         (*insvqi_1_mem_rex64): Ditto.
42380         (*insvqi_2): Ditto.
42381         (*insvqi_3): Rename from *insvqi.
42383         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42385 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42387         * doc/cfg.texi (Edges): Remove reference to Java.
42388         (Maintaining the CFG): Ditto.
42390 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42392         PR middle-end/77674
42393         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42394         transparent aliases.
42396 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42398         PR middle-end/77484
42399         * predict.def (PRED_CALL): Update hitrate.
42400         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42401         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42402         into direct/indirect/polymorphic variants.
42404 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42406         Update copyright years.
42408         * gcc.c (process_command): Update copyright notice dates.
42409         * gcov-dump.c (print_version): Ditto.
42410         * gcov.c (print_version): Ditto.
42411         * gcov-tool.c (print_version): Ditto.
42412         * gengtype.c (create_file): Ditto.
42413         * doc/cpp.texi: Bump @copying's copyright year.
42414         * doc/cppinternals.texi: Ditto.
42415         * doc/gcc.texi: Ditto.
42416         * doc/gccint.texi: Ditto.
42417         * doc/gcov.texi: Ditto.
42418         * doc/install.texi: Ditto.
42419         * doc/invoke.texi: Ditto.
42421 Copyright (C) 2017 Free Software Foundation, Inc.
42423 Copying and distribution of this file, with or without modification,
42424 are permitted in any medium without royalty provided the copyright
42425 notice and this notice are preserved.